diff --git a/main/404.html b/main/404.html index befeed2d..d3c69fbc 100644 --- a/main/404.html +++ b/main/404.html @@ -56,6 +56,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/LICENSE.html b/main/LICENSE.html index 5841aed6..740a9d4b 100644 --- a/main/LICENSE.html +++ b/main/LICENSE.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/articles/CondMean_Inference.html b/main/articles/CondMean_Inference.html index 5d677a87..76d84c6a 100644 --- a/main/articles/CondMean_Inference.html +++ b/main/articles/CondMean_Inference.html @@ -58,6 +58,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/articles/quickstart.html b/main/articles/quickstart.html index 66db78c4..986b7ff2 100644 --- a/main/articles/quickstart.html +++ b/main/articles/quickstart.html @@ -58,6 +58,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/index.html b/main/index.html index 92f19e69..f4d0cebc 100644 --- a/main/index.html +++ b/main/index.html @@ -68,6 +68,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/QR_decomp.html b/main/reference/QR_decomp.html index ead80b04..75e769d5 100644 --- a/main/reference/QR_decomp.html +++ b/main/reference/QR_decomp.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/Stack.html b/main/reference/Stack.html index b74123ff..5e909e38 100644 --- a/main/reference/Stack.html +++ b/main/reference/Stack.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/add_class.html b/main/reference/add_class.html index a417e5c2..3a457076 100644 --- a/main/reference/add_class.html +++ b/main/reference/add_class.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/adjust_trajectories.html b/main/reference/adjust_trajectories.html index 2257bb85..62dfd222 100644 --- a/main/reference/adjust_trajectories.html +++ b/main/reference/adjust_trajectories.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/adjust_trajectories_single.html b/main/reference/adjust_trajectories_single.html index bf89df54..b9f6b1a8 100644 --- a/main/reference/adjust_trajectories_single.html +++ b/main/reference/adjust_trajectories_single.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/analyse.html b/main/reference/analyse.html index b135806e..cabd569a 100644 --- a/main/reference/analyse.html +++ b/main/reference/analyse.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/ancova.html b/main/reference/ancova.html index 810f9b5a..35620fc7 100644 --- a/main/reference/ancova.html +++ b/main/reference/ancova.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/ancova_single.html b/main/reference/ancova_single.html index 78dca8bd..35354c27 100644 --- a/main/reference/ancova_single.html +++ b/main/reference/ancova_single.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/antidepressant_data.html b/main/reference/antidepressant_data.html index 8ff765ed..b41ce592 100644 --- a/main/reference/antidepressant_data.html +++ b/main/reference/antidepressant_data.html @@ -52,6 +52,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/apply_delta.html b/main/reference/apply_delta.html index f99b7f5f..47c20b29 100644 --- a/main/reference/apply_delta.html +++ b/main/reference/apply_delta.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/as_analysis.html b/main/reference/as_analysis.html index f3d454f9..9332b790 100644 --- a/main/reference/as_analysis.html +++ b/main/reference/as_analysis.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/as_ascii_table.html b/main/reference/as_ascii_table.html index dd5a6f18..711609df 100644 --- a/main/reference/as_ascii_table.html +++ b/main/reference/as_ascii_table.html @@ -40,6 +40,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/as_class.html b/main/reference/as_class.html index cd2f6b00..2d92c233 100644 --- a/main/reference/as_class.html +++ b/main/reference/as_class.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/as_cropped_char.html b/main/reference/as_cropped_char.html index ac98eda0..73cc16a6 100644 --- a/main/reference/as_cropped_char.html +++ b/main/reference/as_cropped_char.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/as_dataframe.html b/main/reference/as_dataframe.html index 6fad37b3..ee50c731 100644 --- a/main/reference/as_dataframe.html +++ b/main/reference/as_dataframe.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/as_draws.html b/main/reference/as_draws.html index d854350e..5f81bcdb 100644 --- a/main/reference/as_draws.html +++ b/main/reference/as_draws.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/as_imputation.html b/main/reference/as_imputation.html index ce92d71b..71530714 100644 --- a/main/reference/as_imputation.html +++ b/main/reference/as_imputation.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/as_indices.html b/main/reference/as_indices.html index 086d8901..c117dbc0 100644 --- a/main/reference/as_indices.html +++ b/main/reference/as_indices.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/as_mmrm_df.html b/main/reference/as_mmrm_df.html index 5ae7ea9b..bc57a171 100644 --- a/main/reference/as_mmrm_df.html +++ b/main/reference/as_mmrm_df.html @@ -50,6 +50,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/as_mmrm_formula.html b/main/reference/as_mmrm_formula.html index 33a20b11..a524d8d3 100644 --- a/main/reference/as_mmrm_formula.html +++ b/main/reference/as_mmrm_formula.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/as_model_df.html b/main/reference/as_model_df.html index bcea6a2e..413df4f4 100644 --- a/main/reference/as_model_df.html +++ b/main/reference/as_model_df.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/as_simple_formula.html b/main/reference/as_simple_formula.html index 4fe666c7..e9907ba9 100644 --- a/main/reference/as_simple_formula.html +++ b/main/reference/as_simple_formula.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/as_stan_array.html b/main/reference/as_stan_array.html index aaeacdd8..0953588f 100644 --- a/main/reference/as_stan_array.html +++ b/main/reference/as_stan_array.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/as_strata.html b/main/reference/as_strata.html index fe0367cd..326cafa6 100644 --- a/main/reference/as_strata.html +++ b/main/reference/as_strata.html @@ -46,6 +46,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/assert_variables_exist.html b/main/reference/assert_variables_exist.html index fd86970b..5251f1dc 100644 --- a/main/reference/assert_variables_exist.html +++ b/main/reference/assert_variables_exist.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/char2fct.html b/main/reference/char2fct.html index f29b4581..def784ef 100644 --- a/main/reference/char2fct.html +++ b/main/reference/char2fct.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/check_ESS.html b/main/reference/check_ESS.html index 69ed60e9..73fb82b9 100644 --- a/main/reference/check_ESS.html +++ b/main/reference/check_ESS.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/check_hmc_diagn.html b/main/reference/check_hmc_diagn.html index 4eee67ba..12fcc6e5 100644 --- a/main/reference/check_hmc_diagn.html +++ b/main/reference/check_hmc_diagn.html @@ -46,6 +46,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/check_mcmc.html b/main/reference/check_mcmc.html index e6019608..fdc24a9b 100644 --- a/main/reference/check_mcmc.html +++ b/main/reference/check_mcmc.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/compute_sigma.html b/main/reference/compute_sigma.html index 611b0278..fa47a7f4 100644 --- a/main/reference/compute_sigma.html +++ b/main/reference/compute_sigma.html @@ -40,6 +40,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/convert_to_imputation_list_df.html b/main/reference/convert_to_imputation_list_df.html index ad2663f8..b3631b4d 100644 --- a/main/reference/convert_to_imputation_list_df.html +++ b/main/reference/convert_to_imputation_list_df.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/d_lagscale.html b/main/reference/d_lagscale.html index 05eacabb..0b3445c9 100644 --- a/main/reference/d_lagscale.html +++ b/main/reference/d_lagscale.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/delta_template.html b/main/reference/delta_template.html index db81669c..b8431367 100644 --- a/main/reference/delta_template.html +++ b/main/reference/delta_template.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/do_not_run.html b/main/reference/do_not_run.html index 4ca581b6..ad063782 100644 --- a/main/reference/do_not_run.html +++ b/main/reference/do_not_run.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/draws.html b/main/reference/draws.html index eddb7c57..3e78ce4b 100644 --- a/main/reference/draws.html +++ b/main/reference/draws.html @@ -64,6 +64,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/encap_get_mmrm_sample.html b/main/reference/encap_get_mmrm_sample.html index efdf437a..8ccacc66 100644 --- a/main/reference/encap_get_mmrm_sample.html +++ b/main/reference/encap_get_mmrm_sample.html @@ -42,6 +42,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/eval_mmrm.html b/main/reference/eval_mmrm.html index 82b86adb..e014fb14 100644 --- a/main/reference/eval_mmrm.html +++ b/main/reference/eval_mmrm.html @@ -44,6 +44,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/expand.html b/main/reference/expand.html index bb33564b..cfcff292 100644 --- a/main/reference/expand.html +++ b/main/reference/expand.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/extract_covariates.html b/main/reference/extract_covariates.html index 1f6ae450..3b9367ae 100644 --- a/main/reference/extract_covariates.html +++ b/main/reference/extract_covariates.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/extract_data_nmar_as_na.html b/main/reference/extract_data_nmar_as_na.html index 0bec12dc..838fcb82 100644 --- a/main/reference/extract_data_nmar_as_na.html +++ b/main/reference/extract_data_nmar_as_na.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/extract_draws.html b/main/reference/extract_draws.html index c1618784..6c8e46e3 100644 --- a/main/reference/extract_draws.html +++ b/main/reference/extract_draws.html @@ -40,6 +40,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/extract_imputed_df.html b/main/reference/extract_imputed_df.html index cf689626..2111f000 100644 --- a/main/reference/extract_imputed_df.html +++ b/main/reference/extract_imputed_df.html @@ -46,6 +46,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/extract_imputed_dfs.html b/main/reference/extract_imputed_dfs.html index 29ad6875..c9342462 100644 --- a/main/reference/extract_imputed_dfs.html +++ b/main/reference/extract_imputed_dfs.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/extract_params.html b/main/reference/extract_params.html index 04f9c469..11c527b8 100644 --- a/main/reference/extract_params.html +++ b/main/reference/extract_params.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/fit_mcmc.html b/main/reference/fit_mcmc.html index 222988ee..aa5032ce 100644 --- a/main/reference/fit_mcmc.html +++ b/main/reference/fit_mcmc.html @@ -44,6 +44,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/fit_mmrm.html b/main/reference/fit_mmrm.html index 1e56e7ef..c32aa378 100644 --- a/main/reference/fit_mmrm.html +++ b/main/reference/fit_mmrm.html @@ -40,6 +40,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/generate_data_single.html b/main/reference/generate_data_single.html index 27462b39..ec9aa63e 100644 --- a/main/reference/generate_data_single.html +++ b/main/reference/generate_data_single.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/getStrategies.html b/main/reference/getStrategies.html index 7948f8ea..77b3fc88 100644 --- a/main/reference/getStrategies.html +++ b/main/reference/getStrategies.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/get_ESS.html b/main/reference/get_ESS.html index a8c4524c..ee65e664 100644 --- a/main/reference/get_ESS.html +++ b/main/reference/get_ESS.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/get_bootstrap_stack.html b/main/reference/get_bootstrap_stack.html index 6ed7c0fb..daf9062c 100644 --- a/main/reference/get_bootstrap_stack.html +++ b/main/reference/get_bootstrap_stack.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/get_cluster.html b/main/reference/get_cluster.html index ef3967e7..bdaab58d 100644 --- a/main/reference/get_cluster.html +++ b/main/reference/get_cluster.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/get_conditional_parameters.html b/main/reference/get_conditional_parameters.html index fd82b89e..180494aa 100644 --- a/main/reference/get_conditional_parameters.html +++ b/main/reference/get_conditional_parameters.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/get_delta_template.html b/main/reference/get_delta_template.html index 1b0ddcd1..c62ca5c4 100644 --- a/main/reference/get_delta_template.html +++ b/main/reference/get_delta_template.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/get_draws_mle.html b/main/reference/get_draws_mle.html index c3b3990b..9e1fdd4e 100644 --- a/main/reference/get_draws_mle.html +++ b/main/reference/get_draws_mle.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/get_ests_bmlmi.html b/main/reference/get_ests_bmlmi.html index aec1bc2b..34c216f3 100644 --- a/main/reference/get_ests_bmlmi.html +++ b/main/reference/get_ests_bmlmi.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/get_example_data.html b/main/reference/get_example_data.html index 42091940..b53cad77 100644 --- a/main/reference/get_example_data.html +++ b/main/reference/get_example_data.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/get_jackknife_stack.html b/main/reference/get_jackknife_stack.html index c88be06d..9a9fe272 100644 --- a/main/reference/get_jackknife_stack.html +++ b/main/reference/get_jackknife_stack.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/get_mmrm_sample.html b/main/reference/get_mmrm_sample.html index 2303bd5f..4d1e7cd4 100644 --- a/main/reference/get_mmrm_sample.html +++ b/main/reference/get_mmrm_sample.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/get_pattern_groups.html b/main/reference/get_pattern_groups.html index 6dfb4133..4c94366d 100644 --- a/main/reference/get_pattern_groups.html +++ b/main/reference/get_pattern_groups.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/get_pattern_groups_unique.html b/main/reference/get_pattern_groups_unique.html index 672c3cc2..7dad00e4 100644 --- a/main/reference/get_pattern_groups_unique.html +++ b/main/reference/get_pattern_groups_unique.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/get_pool_components.html b/main/reference/get_pool_components.html index 6ff9f22f..58d4b3cb 100644 --- a/main/reference/get_pool_components.html +++ b/main/reference/get_pool_components.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/get_visit_distribution_parameters.html b/main/reference/get_visit_distribution_parameters.html index 1c6abf2a..ca0b75d9 100644 --- a/main/reference/get_visit_distribution_parameters.html +++ b/main/reference/get_visit_distribution_parameters.html @@ -42,6 +42,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/has_class.html b/main/reference/has_class.html index 32e07e55..e22a8ee4 100644 --- a/main/reference/has_class.html +++ b/main/reference/has_class.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/ife.html b/main/reference/ife.html index 29f257ea..7559f78d 100644 --- a/main/reference/ife.html +++ b/main/reference/ife.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/imputation_df.html b/main/reference/imputation_df.html index 6bdde1bc..07290efe 100644 --- a/main/reference/imputation_df.html +++ b/main/reference/imputation_df.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/imputation_list_df.html b/main/reference/imputation_list_df.html index d235676a..af6266f4 100644 --- a/main/reference/imputation_list_df.html +++ b/main/reference/imputation_list_df.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/imputation_list_single.html b/main/reference/imputation_list_single.html index e6842c61..133e54e8 100644 --- a/main/reference/imputation_list_single.html +++ b/main/reference/imputation_list_single.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/imputation_single.html b/main/reference/imputation_single.html index c51760dc..3b624828 100644 --- a/main/reference/imputation_single.html +++ b/main/reference/imputation_single.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/impute.html b/main/reference/impute.html index 3fad5901..c4b6c6fe 100644 --- a/main/reference/impute.html +++ b/main/reference/impute.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/impute_data_individual.html b/main/reference/impute_data_individual.html index 66fb5e59..6aa04ace 100644 --- a/main/reference/impute_data_individual.html +++ b/main/reference/impute_data_individual.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/impute_internal.html b/main/reference/impute_internal.html index 81f99d22..39a28bd0 100644 --- a/main/reference/impute_internal.html +++ b/main/reference/impute_internal.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/impute_outcome.html b/main/reference/impute_outcome.html index c3a48c47..2fa4a576 100644 --- a/main/reference/impute_outcome.html +++ b/main/reference/impute_outcome.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/index.html b/main/reference/index.html index 9ea69f27..bef92102 100644 --- a/main/reference/index.html +++ b/main/reference/index.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/invert.html b/main/reference/invert.html index 247abc76..99b6b102 100644 --- a/main/reference/invert.html +++ b/main/reference/invert.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/invert_indexes.html b/main/reference/invert_indexes.html index 401d2bb1..b3f67d7a 100644 --- a/main/reference/invert_indexes.html +++ b/main/reference/invert_indexes.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/is_absent.html b/main/reference/is_absent.html index 89043fc4..b535247e 100644 --- a/main/reference/is_absent.html +++ b/main/reference/is_absent.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/is_char_fact.html b/main/reference/is_char_fact.html index 943c934d..b8696126 100644 --- a/main/reference/is_char_fact.html +++ b/main/reference/is_char_fact.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/is_char_one.html b/main/reference/is_char_one.html index f36e770e..0fbf2769 100644 --- a/main/reference/is_char_one.html +++ b/main/reference/is_char_one.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/is_in_rbmi_development.html b/main/reference/is_in_rbmi_development.html index 1c91589a..72b887eb 100644 --- a/main/reference/is_in_rbmi_development.html +++ b/main/reference/is_in_rbmi_development.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/is_num_char_fact.html b/main/reference/is_num_char_fact.html index 7cbfb20a..dd3938e9 100644 --- a/main/reference/is_num_char_fact.html +++ b/main/reference/is_num_char_fact.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/locf.html b/main/reference/locf.html index dedab28f..a7446bf2 100644 --- a/main/reference/locf.html +++ b/main/reference/locf.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/longDataConstructor.html b/main/reference/longDataConstructor.html index 73d0c1a0..34959c23 100644 --- a/main/reference/longDataConstructor.html +++ b/main/reference/longDataConstructor.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/ls_design.html b/main/reference/ls_design.html index f213a580..b22f0ee9 100644 --- a/main/reference/ls_design.html +++ b/main/reference/ls_design.html @@ -44,6 +44,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/lsmeans.html b/main/reference/lsmeans.html index a6be5c78..058d2f76 100644 --- a/main/reference/lsmeans.html +++ b/main/reference/lsmeans.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/method.html b/main/reference/method.html index 7dc2c32f..a86a5c31 100644 --- a/main/reference/method.html +++ b/main/reference/method.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/parametric_ci.html b/main/reference/parametric_ci.html index 1fcf0092..a0e53e25 100644 --- a/main/reference/parametric_ci.html +++ b/main/reference/parametric_ci.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/pool.html b/main/reference/pool.html index 14be8d30..5346890b 100644 --- a/main/reference/pool.html +++ b/main/reference/pool.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/pool_bootstrap_normal.html b/main/reference/pool_bootstrap_normal.html index 7f57e638..0ccd5089 100644 --- a/main/reference/pool_bootstrap_normal.html +++ b/main/reference/pool_bootstrap_normal.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/pool_bootstrap_percentile.html b/main/reference/pool_bootstrap_percentile.html index 5f57d460..1f225de4 100644 --- a/main/reference/pool_bootstrap_percentile.html +++ b/main/reference/pool_bootstrap_percentile.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/pool_internal.html b/main/reference/pool_internal.html index a3ae1388..2f14f936 100644 --- a/main/reference/pool_internal.html +++ b/main/reference/pool_internal.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/prepare_stan_data.html b/main/reference/prepare_stan_data.html index 521301b3..c52729fc 100644 --- a/main/reference/prepare_stan_data.html +++ b/main/reference/prepare_stan_data.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/print.analysis.html b/main/reference/print.analysis.html index b7c47b87..92e99cb6 100644 --- a/main/reference/print.analysis.html +++ b/main/reference/print.analysis.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/print.draws.html b/main/reference/print.draws.html index 4fdbb5e0..8f2d4569 100644 --- a/main/reference/print.draws.html +++ b/main/reference/print.draws.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/print.imputation.html b/main/reference/print.imputation.html index 9dfe59c6..915d23ee 100644 --- a/main/reference/print.imputation.html +++ b/main/reference/print.imputation.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/progressLogger.html b/main/reference/progressLogger.html index 972922cd..a20712a2 100644 --- a/main/reference/progressLogger.html +++ b/main/reference/progressLogger.html @@ -42,6 +42,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/pval_percentile.html b/main/reference/pval_percentile.html index 290d85c4..e972b942 100644 --- a/main/reference/pval_percentile.html +++ b/main/reference/pval_percentile.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/random_effects_expr.html b/main/reference/random_effects_expr.html index 3eb286f5..66965a22 100644 --- a/main/reference/random_effects_expr.html +++ b/main/reference/random_effects_expr.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/rbmi-package.html b/main/reference/rbmi-package.html index 08134cf8..ac3326b8 100644 --- a/main/reference/rbmi-package.html +++ b/main/reference/rbmi-package.html @@ -56,6 +56,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/record.html b/main/reference/record.html index 543a1cdc..bffc8e1e 100644 --- a/main/reference/record.html +++ b/main/reference/record.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/recursive_reduce.html b/main/reference/recursive_reduce.html index 615131bd..5cecab0e 100644 --- a/main/reference/recursive_reduce.html +++ b/main/reference/recursive_reduce.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/remove_if_all_missing.html b/main/reference/remove_if_all_missing.html index ce349e04..c63dbf49 100644 --- a/main/reference/remove_if_all_missing.html +++ b/main/reference/remove_if_all_missing.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/rubin_df.html b/main/reference/rubin_df.html index 60eac45b..36cbe8f9 100644 --- a/main/reference/rubin_df.html +++ b/main/reference/rubin_df.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/rubin_rules.html b/main/reference/rubin_rules.html index 4b7ea74f..afee94e0 100644 --- a/main/reference/rubin_rules.html +++ b/main/reference/rubin_rules.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/sample_ids.html b/main/reference/sample_ids.html index 167bc249..323e0ebb 100644 --- a/main/reference/sample_ids.html +++ b/main/reference/sample_ids.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/sample_list.html b/main/reference/sample_list.html index 6a6f0dc5..22d3bff5 100644 --- a/main/reference/sample_list.html +++ b/main/reference/sample_list.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/sample_mvnorm.html b/main/reference/sample_mvnorm.html index f8a1c152..321abd3e 100644 --- a/main/reference/sample_mvnorm.html +++ b/main/reference/sample_mvnorm.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/sample_single.html b/main/reference/sample_single.html index 945f5bda..c4b1a0fb 100644 --- a/main/reference/sample_single.html +++ b/main/reference/sample_single.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/scalerConstructor.html b/main/reference/scalerConstructor.html index a6fd33b9..d7d181c7 100644 --- a/main/reference/scalerConstructor.html +++ b/main/reference/scalerConstructor.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/set_simul_pars.html b/main/reference/set_simul_pars.html index bb13ee39..2d639728 100644 --- a/main/reference/set_simul_pars.html +++ b/main/reference/set_simul_pars.html @@ -46,6 +46,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/set_vars.html b/main/reference/set_vars.html index 5aa1028a..acee9ce5 100644 --- a/main/reference/set_vars.html +++ b/main/reference/set_vars.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/simulate_data.html b/main/reference/simulate_data.html index 4b13d40a..b23c4117 100644 --- a/main/reference/simulate_data.html +++ b/main/reference/simulate_data.html @@ -46,6 +46,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/simulate_dropout.html b/main/reference/simulate_dropout.html index 39aa1a6a..003dbaf8 100644 --- a/main/reference/simulate_dropout.html +++ b/main/reference/simulate_dropout.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/simulate_ice.html b/main/reference/simulate_ice.html index b20452ef..26df0376 100644 --- a/main/reference/simulate_ice.html +++ b/main/reference/simulate_ice.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/simulate_test_data.html b/main/reference/simulate_test_data.html index 0849442c..f0841a33 100644 --- a/main/reference/simulate_test_data.html +++ b/main/reference/simulate_test_data.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/sort_by.html b/main/reference/sort_by.html index 69f150ea..b0e2b801 100644 --- a/main/reference/sort_by.html +++ b/main/reference/sort_by.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/split_dim.html b/main/reference/split_dim.html index 549f1cce..5f8aa782 100644 --- a/main/reference/split_dim.html +++ b/main/reference/split_dim.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/split_imputations.html b/main/reference/split_imputations.html index 5ca59ead..033e395f 100644 --- a/main/reference/split_imputations.html +++ b/main/reference/split_imputations.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/str_contains.html b/main/reference/str_contains.html index 3dcd82c0..2ef5f962 100644 --- a/main/reference/str_contains.html +++ b/main/reference/str_contains.html @@ -44,6 +44,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/strategies.html b/main/reference/strategies.html index 32a7b301..7e39ff1e 100644 --- a/main/reference/strategies.html +++ b/main/reference/strategies.html @@ -40,6 +40,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/string_pad.html b/main/reference/string_pad.html index 02fc2d30..6b71fd71 100644 --- a/main/reference/string_pad.html +++ b/main/reference/string_pad.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/transpose_imputations.html b/main/reference/transpose_imputations.html index 808bb817..78b2cbd2 100644 --- a/main/reference/transpose_imputations.html +++ b/main/reference/transpose_imputations.html @@ -46,6 +46,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/transpose_results.html b/main/reference/transpose_results.html index 0e3090ff..52b0f46b 100644 --- a/main/reference/transpose_results.html +++ b/main/reference/transpose_results.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/transpose_samples.html b/main/reference/transpose_samples.html index 91305625..740226c1 100644 --- a/main/reference/transpose_samples.html +++ b/main/reference/transpose_samples.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/validate.analysis.html b/main/reference/validate.analysis.html index 26b9dc54..d2d8fcbb 100644 --- a/main/reference/validate.analysis.html +++ b/main/reference/validate.analysis.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/validate.draws.html b/main/reference/validate.draws.html index e3bd8079..2323cdea 100644 --- a/main/reference/validate.draws.html +++ b/main/reference/validate.draws.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/validate.html b/main/reference/validate.html index b76b804e..63e64c91 100644 --- a/main/reference/validate.html +++ b/main/reference/validate.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/validate.is_mar.html b/main/reference/validate.is_mar.html index 4d98247f..c02cfc65 100644 --- a/main/reference/validate.is_mar.html +++ b/main/reference/validate.is_mar.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/validate.ivars.html b/main/reference/validate.ivars.html index 90ff4d44..f084a6e1 100644 --- a/main/reference/validate.ivars.html +++ b/main/reference/validate.ivars.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/validate.references.html b/main/reference/validate.references.html index 47231854..13818ad2 100644 --- a/main/reference/validate.references.html +++ b/main/reference/validate.references.html @@ -36,6 +36,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/validate.sample_list.html b/main/reference/validate.sample_list.html index 6c995457..7fd71c58 100644 --- a/main/reference/validate.sample_list.html +++ b/main/reference/validate.sample_list.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/validate.sample_single.html b/main/reference/validate.sample_single.html index fa7ddba4..d62e71ac 100644 --- a/main/reference/validate.sample_single.html +++ b/main/reference/validate.sample_single.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/validate.simul_pars.html b/main/reference/validate.simul_pars.html index 3546ee68..8d5fbd8e 100644 --- a/main/reference/validate.simul_pars.html +++ b/main/reference/validate.simul_pars.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/validate.stan_data.html b/main/reference/validate.stan_data.html index c921267e..929b8e47 100644 --- a/main/reference/validate.stan_data.html +++ b/main/reference/validate.stan_data.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/validate_analyse_pars.html b/main/reference/validate_analyse_pars.html index a842c947..874ea49a 100644 --- a/main/reference/validate_analyse_pars.html +++ b/main/reference/validate_analyse_pars.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/validate_datalong.html b/main/reference/validate_datalong.html index 5435f30c..b28bf61a 100644 --- a/main/reference/validate_datalong.html +++ b/main/reference/validate_datalong.html @@ -34,6 +34,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/reference/validate_strategies.html b/main/reference/validate_strategies.html index 2ec7bd01..dc6b8882 100644 --- a/main/reference/validate_strategies.html +++ b/main/reference/validate_strategies.html @@ -38,6 +38,12 @@
  • Changelog
    • +
  • +
    main
    • + diff --git a/main/search.json b/main/search.json index 77cefd44..340b55ec 100644 --- a/main/search.json +++ b/main/search.json @@ -1 +1 @@ -[{"path":"/CONTRIBUTING.html","id":null,"dir":"","previous_headings":"","what":"Contributing to rbmi","title":"Contributing to rbmi","text":"file outlines propose make changes rbmi well providing details obscure aspects package’s development process.","code":""},{"path":"/CONTRIBUTING.html","id":"setup","dir":"","previous_headings":"","what":"Setup","title":"Contributing to rbmi","text":"order develop contribute rbmi need access C/C++ compiler. Windows install rtools macOS install Xcode. Likewise, also need install package’s development dependencies. can done launching R within project root executing:","code":"devtools::install_dev_deps()"},{"path":"/CONTRIBUTING.html","id":"code-changes","dir":"","previous_headings":"","what":"Code changes","title":"Contributing to rbmi","text":"want make code contribution, ’s good idea first file issue make sure someone team agrees ’s needed. ’ve found bug, please file issue illustrates bug minimal reprex (also help write unit test, needed).","code":""},{"path":"/CONTRIBUTING.html","id":"pull-request-process","dir":"","previous_headings":"Code changes","what":"Pull request process","title":"Contributing to rbmi","text":"project uses simple GitHub flow model development. , code changes done feature branch based main branch merged back main branch complete. Pull Requests accepted unless CI/CD checks passed. (See CI/CD section information). Pull Requests relating package’s core R code must accompanied corresponding unit test. pull requests containing changes core R code contain unit test demonstrate working intended accepted. (See Unit Testing section information). Pull Requests add lines changed NEWS.md file.","code":""},{"path":"/CONTRIBUTING.html","id":"coding-considerations","dir":"","previous_headings":"Code changes","what":"Coding Considerations","title":"Contributing to rbmi","text":"use roxygen2, Markdown syntax, documentation. Please ensure code conforms lintr. can check running lintr::lint(\"FILE NAME\") files modified ensuring findings kept possible. hard requirements following lintr’s conventions encourage developers follow guidance closely possible. project uses 4 space indents, contributions following accepted. project makes use S3 R6 OOP. Usage S4 OOP systems avoided unless absolutely necessary ensure consistency. said recommended stick S3 unless modification place R6 specific features required. current desire package keep dependency tree small possible. end discouraged adding additional packages “Depends” / “Imports” section unless absolutely essential. importing package just use single function consider just copying source code function instead, though please check licence include proper attribution/notices. expectations “Suggests” free use package vignettes / unit tests, though please mindful unnecessarily excessive .","code":""},{"path":"/CONTRIBUTING.html","id":"unit-testing--cicd","dir":"","previous_headings":"","what":"Unit Testing & CI/CD","title":"Contributing to rbmi","text":"project uses testthat perform unit testing combination GitHub Actions CI/CD.","code":""},{"path":"/CONTRIBUTING.html","id":"scheduled-testing","dir":"","previous_headings":"Unit Testing & CI/CD","what":"Scheduled Testing","title":"Contributing to rbmi","text":"Due stochastic nature package unit tests take considerable amount time execute. avoid issues usability, unit tests take couple seconds run deferred scheduled testing. tests run occasionally periodic basis (currently twice month) every pull request / push event. defer test scheduled build simply include skip_if_not(is_full_test()) top test_that() block .e. scheduled tests can also manually activated going “https://github.com/insightsengineering/rbmi” -> “Actions” -> “Bi-Weekly” -> “Run Workflow”. advisable releasing CRAN.","code":"test_that(\"some unit test\", { skip_if_not(is_full_test()) expect_equal(1,1) })"},{"path":"/CONTRIBUTING.html","id":"cran-releases","dir":"","previous_headings":"Unit Testing & CI/CD","what":"CRAN Releases","title":"Contributing to rbmi","text":"order release package CRAN needs tested across multiple different OS’s versions R. implemented project via GitHub Action Workflow titled “Check CRAN” needs manually activated. go “https://github.com/insightsengineering/rbmi” -> “Actions” -> “Check CRAN” -> “Run Workflow”. tests pass package can safely released CRAN (updating relevant cran-comments.md file)","code":""},{"path":"/CONTRIBUTING.html","id":"docker-images","dir":"","previous_headings":"Unit Testing & CI/CD","what":"Docker Images","title":"Contributing to rbmi","text":"support CI/CD terms reducing installation time, several Docker images pre-built contain packages system dependencies project needs. current relevant images can found : ghcr.io/insightsengineering/rbmi:r404 ghcr.io/insightsengineering/rbmi:r410 ghcr.io/insightsengineering/rbmi:latest latest image automatically re-built month contain latest version R packages. versions built older versions R (indicated tag number) contain package versions version R released. important ensure package works older versions R many companies typically run due delays validation processes. code create images can found misc/docker. legacy images (.e. everything excluding “latest” image) built manual request running corresponding GitHub Actions Workflow.","code":""},{"path":"/CONTRIBUTING.html","id":"reproducibility-print-tests--snaps","dir":"","previous_headings":"Unit Testing & CI/CD","what":"Reproducibility, Print Tests & Snaps","title":"Contributing to rbmi","text":"particular issue testing package reproducibility. part handled well via set.seed() however stan/rstan guarantee reproducibility even seed run different hardware. issue surfaces testing print messages pool object displays treatment estimates thus identical run different machines. address issue pre-made pool objects generated stored R/sysdata.rda (generated data-raw/create_print_test_data.R). generated print messages compared expected values stored tests/testthat/_snaps/ (automatically created testthat::expect_snapshot())","code":""},{"path":"/CONTRIBUTING.html","id":"fitting-mmrms","dir":"","previous_headings":"","what":"Fitting MMRM’s","title":"Contributing to rbmi","text":"package currently uses mmrm package fit MMRM models. package still fairly new far proven stable, fast reliable. spot issues MMRM package please raise corresponding GitHub Repository - link mmrm package uses TMB uncommon see warnings either inconsistent versions TMB Matrix package compiled . order resolve may wish re-compile packages source using: Note need rtools installed Windows machine Xcode running macOS (somehow else access C/C++ compiler).","code":"install.packages(c(\"TMB\", \"mmrm\"), type = \"source\")"},{"path":"/CONTRIBUTING.html","id":"rstan","dir":"","previous_headings":"","what":"rstan","title":"Contributing to rbmi","text":"Bayesian models fitted package implemented via stan/rstan. code can found inst/stan/MMRM.stan. Note package automatically take care compiling code install run devtools::load_all(). Please note package won’t recompile code unless changed source code delete src directory.","code":""},{"path":"/CONTRIBUTING.html","id":"vignettes","dir":"","previous_headings":"","what":"Vignettes","title":"Contributing to rbmi","text":"CRAN imposes 10-minute run limit building, compiling testing package. keep limit vignettes pre-built; say simply changing source code automatically update vignettes, need manually re-build . need run: re-built need commit updated *.html files git repository. reference static vignette process works using “asis” vignette engine provided R.rsp. works getting R recognise vignettes files ending *.html.asis; builds simply copying corresponding files ending *.html relevent docs/ folder built package.","code":"Rscript vignettes/build.R"},{"path":"/CONTRIBUTING.html","id":"misc--local-folders","dir":"","previous_headings":"","what":"Misc & Local Folders","title":"Contributing to rbmi","text":"misc/ folder project used hold useful scripts, analyses, simulations & infrastructure code wish keep isn’t essential build deployment package. Feel free store additional stuff feel worth keeping. Likewise, local/ added .gitignore file meaning anything stored folder won’t committed repository. example, may find useful storing personal scripts testing generally exploring package development.","code":""},{"path":"/LICENSE.html","id":null,"dir":"","previous_headings":"","what":"Apache License","title":"Apache License","text":"Version 2.0, January 2004 ","code":""},{"path":[]},{"path":"/LICENSE.html","id":"id_1-definitions","dir":"","previous_headings":"Terms and Conditions for use, reproduction, and distribution","what":"1. Definitions","title":"Apache License","text":"“License” shall mean terms conditions use, reproduction, distribution defined Sections 1 9 document. “Licensor” shall mean copyright owner entity authorized copyright owner granting License. “Legal Entity” shall mean union acting entity entities control, controlled , common control entity. purposes definition, “control” means () power, direct indirect, cause direction management entity, whether contract otherwise, (ii) ownership fifty percent (50%) outstanding shares, (iii) beneficial ownership entity. “” (“”) shall mean individual Legal Entity exercising permissions granted License. “Source” form shall mean preferred form making modifications, including limited software source code, documentation source, configuration files. “Object” form shall mean form resulting mechanical transformation translation Source form, including limited compiled object code, generated documentation, conversions media types. “Work” shall mean work authorship, whether Source Object form, made available License, indicated copyright notice included attached work (example provided Appendix ). “Derivative Works” shall mean work, whether Source Object form, based (derived ) Work editorial revisions, annotations, elaborations, modifications represent, whole, original work authorship. purposes License, Derivative Works shall include works remain separable , merely link (bind name) interfaces , Work Derivative Works thereof. “Contribution” shall mean work authorship, including original version Work modifications additions Work Derivative Works thereof, intentionally submitted Licensor inclusion Work copyright owner individual Legal Entity authorized submit behalf copyright owner. purposes definition, “submitted” means form electronic, verbal, written communication sent Licensor representatives, including limited communication electronic mailing lists, source code control systems, issue tracking systems managed , behalf , Licensor purpose discussing improving Work, excluding communication conspicuously marked otherwise designated writing copyright owner “Contribution.” “Contributor” shall mean Licensor individual Legal Entity behalf Contribution received Licensor subsequently incorporated within Work.","code":""},{"path":"/LICENSE.html","id":"id_2-grant-of-copyright-license","dir":"","previous_headings":"Terms and Conditions for use, reproduction, and distribution","what":"2. Grant of Copyright License","title":"Apache License","text":"Subject terms conditions License, Contributor hereby grants perpetual, worldwide, non-exclusive, -charge, royalty-free, irrevocable copyright license reproduce, prepare Derivative Works , publicly display, publicly perform, sublicense, distribute Work Derivative Works Source Object form.","code":""},{"path":"/LICENSE.html","id":"id_3-grant-of-patent-license","dir":"","previous_headings":"Terms and Conditions for use, reproduction, and distribution","what":"3. Grant of Patent License","title":"Apache License","text":"Subject terms conditions License, Contributor hereby grants perpetual, worldwide, non-exclusive, -charge, royalty-free, irrevocable (except stated section) patent license make, made, use, offer sell, sell, import, otherwise transfer Work, license applies patent claims licensable Contributor necessarily infringed Contribution(s) alone combination Contribution(s) Work Contribution(s) submitted. institute patent litigation entity (including cross-claim counterclaim lawsuit) alleging Work Contribution incorporated within Work constitutes direct contributory patent infringement, patent licenses granted License Work shall terminate date litigation filed.","code":""},{"path":"/LICENSE.html","id":"id_4-redistribution","dir":"","previous_headings":"Terms and Conditions for use, reproduction, and distribution","what":"4. Redistribution","title":"Apache License","text":"may reproduce distribute copies Work Derivative Works thereof medium, without modifications, Source Object form, provided meet following conditions: () must give recipients Work Derivative Works copy License; (b) must cause modified files carry prominent notices stating changed files; (c) must retain, Source form Derivative Works distribute, copyright, patent, trademark, attribution notices Source form Work, excluding notices pertain part Derivative Works; (d) Work includes “NOTICE” text file part distribution, Derivative Works distribute must include readable copy attribution notices contained within NOTICE file, excluding notices pertain part Derivative Works, least one following places: within NOTICE text file distributed part Derivative Works; within Source form documentation, provided along Derivative Works; , within display generated Derivative Works, wherever third-party notices normally appear. contents NOTICE file informational purposes modify License. may add attribution notices within Derivative Works distribute, alongside addendum NOTICE text Work, provided additional attribution notices construed modifying License. may add copyright statement modifications may provide additional different license terms conditions use, reproduction, distribution modifications, Derivative Works whole, provided use, reproduction, distribution Work otherwise complies conditions stated License.","code":""},{"path":"/LICENSE.html","id":"id_5-submission-of-contributions","dir":"","previous_headings":"Terms and Conditions for use, reproduction, and distribution","what":"5. Submission of Contributions","title":"Apache License","text":"Unless explicitly state otherwise, Contribution intentionally submitted inclusion Work Licensor shall terms conditions License, without additional terms conditions. Notwithstanding , nothing herein shall supersede modify terms separate license agreement may executed Licensor regarding Contributions.","code":""},{"path":"/LICENSE.html","id":"id_6-trademarks","dir":"","previous_headings":"Terms and Conditions for use, reproduction, and distribution","what":"6. Trademarks","title":"Apache License","text":"License grant permission use trade names, trademarks, service marks, product names Licensor, except required reasonable customary use describing origin Work reproducing content NOTICE file.","code":""},{"path":"/LICENSE.html","id":"id_7-disclaimer-of-warranty","dir":"","previous_headings":"Terms and Conditions for use, reproduction, and distribution","what":"7. Disclaimer of Warranty","title":"Apache License","text":"Unless required applicable law agreed writing, Licensor provides Work (Contributor provides Contributions) “” BASIS, WITHOUT WARRANTIES CONDITIONS KIND, either express implied, including, without limitation, warranties conditions TITLE, NON-INFRINGEMENT, MERCHANTABILITY, FITNESS PARTICULAR PURPOSE. solely responsible determining appropriateness using redistributing Work assume risks associated exercise permissions License.","code":""},{"path":"/LICENSE.html","id":"id_8-limitation-of-liability","dir":"","previous_headings":"Terms and Conditions for use, reproduction, and distribution","what":"8. Limitation of Liability","title":"Apache License","text":"event legal theory, whether tort (including negligence), contract, otherwise, unless required applicable law (deliberate grossly negligent acts) agreed writing, shall Contributor liable damages, including direct, indirect, special, incidental, consequential damages character arising result License use inability use Work (including limited damages loss goodwill, work stoppage, computer failure malfunction, commercial damages losses), even Contributor advised possibility damages.","code":""},{"path":"/LICENSE.html","id":"id_9-accepting-warranty-or-additional-liability","dir":"","previous_headings":"Terms and Conditions for use, reproduction, and distribution","what":"9. Accepting Warranty or Additional Liability","title":"Apache License","text":"redistributing Work Derivative Works thereof, may choose offer, charge fee , acceptance support, warranty, indemnity, liability obligations /rights consistent License. However, accepting obligations, may act behalf sole responsibility, behalf Contributor, agree indemnify, defend, hold Contributor harmless liability incurred , claims asserted , Contributor reason accepting warranty additional liability. END TERMS CONDITIONS","code":""},{"path":"/LICENSE.html","id":"appendix-how-to-apply-the-apache-license-to-your-work","dir":"","previous_headings":"","what":"APPENDIX: How to apply the Apache License to your work","title":"Apache License","text":"apply Apache License work, attach following boilerplate notice, fields enclosed brackets [] replaced identifying information. (Don’t include brackets!) text enclosed appropriate comment syntax file format. also recommend file class name description purpose included “printed page” copyright notice easier identification within third-party archives.","code":"Copyright [yyyy] [name of copyright owner] Licensed under the Apache License, Version 2.0 (the \"License\"); you may not use this file except in compliance with the License. You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0 Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License."},{"path":"/articles/CondMean_Inference.html","id":"introduction","dir":"Articles","previous_headings":"","what":"Introduction","title":"rbmi: Inference with Conditional Mean Imputation","text":"described section 3.10.2 statistical specifications package (vignette(topic = \"stat_specs\", package = \"rbmi\")), two different types variance estimators proposed reference-based imputation methods statistical literature (Bartlett (2023)). first frequentist variance describes actual repeated sampling variability estimator results inference correct frequentist sense, .e. hypothesis tests accurate type error control confidence intervals correct coverage probabilities repeated sampling reference-based assumption correctly specified (Bartlett (2023), Wolbers et al. (2022)). Reference-based missing data assumption strong borrow information control arm imputation active arm. consequence, size frequentist standard errors treatment effects may decrease increasing amounts missing data. second -called “information-anchored” variance originally proposed context sensitivity analyses (Cro, Carpenter, Kenward (2019)). variance estimator based disentangling point estimation variance estimation altogether. resulting information-anchored variance typically similar variance missing--random (MAR) imputation increases increasing amounts missing data approximately rate MAR imputation. However, information-anchored variance reflect actual variability reference-based estimator resulting frequentist inference highly conservative resulting substantial power loss. Reference-based conditional mean imputation combined resampling method jackknife bootstrap first introduced Wolbers et al. (2022). approach naturally targets frequentist variance. information-anchored variance typically estimated using Rubin’s rules Bayesian multiple imputation applicable within conditional mean imputation framework. However, alternative information-anchored variance proposed Lu (2021) can easily obtained show . basic idea Lu (2021) obtain information-anchored variance via MAR imputation combined delta-adjustment delta selected data-driven way match reference-based estimator. conditional mean imputation, proposal Lu (2021) can implemented choosing delta-adjustment difference conditional mean imputation chosen reference-based assumption MAR original dataset. variance can obtained via jackknife bootstrap keeping delta-adjustment fixed. resulting variance estimate similar Rubin’s variance. Moreover, shown Cro, Carpenter, Kenward (2019), variance MAR-imputation combined delta-adjustment achieves even better information-anchoring properties Rubin’s variance reference-based imputation. Reference-based missing data assumptions strong borrow information control arm imputation active arm. vignette demonstrates first obtain frequentist inference using reference-based conditional mean imputation using rbmi, shows information-anchored inference can also easily implemented using package.","code":""},{"path":"/articles/CondMean_Inference.html","id":"data-and-model-specification","dir":"Articles","previous_headings":"","what":"Data and model specification","title":"rbmi: Inference with Conditional Mean Imputation","text":"use publicly available example dataset antidepressant clinical trial active drug versus placebo. relevant endpoint Hamilton 17-item depression rating scale (HAMD17) assessed baseline weeks 1, 2, 4, 6. Study drug discontinuation occurred 24% subjects active drug 26% subjects placebo. data study drug discontinuation missing single additional intermittent missing observation. consider imputation model mean change baseline HAMD17 score outcome (variable CHANGE dataset). following covariates included imputation model: treatment group (THERAPY), (categorical) visit (VISIT), treatment--visit interactions, baseline HAMD17 score (BASVAL), baseline HAMD17 score--visit interactions. common unstructured covariance matrix structure assumed groups. analysis model ANCOVA model treatment group primary factor adjustment baseline HAMD17 score. example, assume imputation strategy ICE “study-drug discontinuation” Jump Reference (JR) subjects imputation based conditional mean imputation combined jackknife resampling (bootstrap also selected).","code":""},{"path":"/articles/CondMean_Inference.html","id":"reference-based-conditional-mean-imputation---frequentist-inference","dir":"Articles","previous_headings":"","what":"Reference-based conditional mean imputation - frequentist inference","title":"rbmi: Inference with Conditional Mean Imputation","text":"Conditional mean imputation combined resampling method jackknife bootstrap naturally targets frequentist estimation standard error treatment effect, thus providing valid frequentist inference. provide code obtain frequentist inference reference-based conditional mean imputation using rbmi. code used section almost identical code quickstart vignette (vignette(topic = \"quickstart\", package = \"rbmi\")) except use conditional mean imputation combined jackknife (method_condmean(type = \"jackknife\")) rather Bayesian multiple imputation (method_bayes()). therefore refer vignette help files individual functions explanations details.","code":""},{"path":"/articles/CondMean_Inference.html","id":"draws","dir":"Articles","previous_headings":"3 Reference-based conditional mean imputation - frequentist inference","what":"Draws","title":"rbmi: Inference with Conditional Mean Imputation","text":"make use rbmi::expand_locf() expand dataset order one row per subject per visit missing outcomes denoted NA. construct data_ice, vars method input arguments first core rbmi function, draws(). Finally, call function draws() derive parameter estimates base imputation model full dataset leave-one-subject-samples.","code":"library(rbmi) library(dplyr) #> #> Attaching package: 'dplyr' #> The following objects are masked from 'package:stats': #> #> filter, lag #> The following objects are masked from 'package:base': #> #> intersect, setdiff, setequal, union dat <- antidepressant_data # Use expand_locf to add rows corresponding to visits with missing outcomes to # the dataset dat <- expand_locf( dat, PATIENT = levels(dat$PATIENT), # expand by PATIENT and VISIT VISIT = levels(dat$VISIT), vars = c(\"BASVAL\", \"THERAPY\"), # fill with LOCF BASVAL and THERAPY group = c(\"PATIENT\"), order = c(\"PATIENT\", \"VISIT\") ) # create data_ice and set the imputation strategy to JR for # each patient with at least one missing observation dat_ice <- dat %>% arrange(PATIENT, VISIT) %>% filter(is.na(CHANGE)) %>% group_by(PATIENT) %>% slice(1) %>% ungroup() %>% select(PATIENT, VISIT) %>% mutate(strategy = \"JR\") # In this dataset, subject 3618 has an intermittent missing values which # does not correspond to a study drug discontinuation. We therefore remove # this subject from `dat_ice`. (In the later imputation step, it will # automatically be imputed under the default MAR assumption.) dat_ice <- dat_ice[-which(dat_ice$PATIENT == 3618),] # Define the names of key variables in our dataset and # the covariates included in the imputation model using `set_vars()` vars <- set_vars( outcome = \"CHANGE\", visit = \"VISIT\", subjid = \"PATIENT\", group = \"THERAPY\", covariates = c(\"BASVAL*VISIT\", \"THERAPY*VISIT\") ) # Define which imputation method to use (here: conditional mean imputation # with jackknife as resampling) method <- method_condmean(type = \"jackknife\") # Create samples for the imputation parameters by running the draws() function drawObj <- draws( data = dat, data_ice = dat_ice, vars = vars, method = method, quiet = TRUE ) drawObj #> #> Draws Object #> ------------ #> Number of Samples: 1 + 172 #> Number of Failed Samples: 0 #> Model Formula: CHANGE ~ 1 + THERAPY + VISIT + BASVAL * VISIT + THERAPY * VISIT #> Imputation Type: condmean #> Method: #> name: Conditional Mean #> covariance: us #> threshold: 0.01 #> same_cov: TRUE #> REML: TRUE #> type: jackknife"},{"path":"/articles/CondMean_Inference.html","id":"impute","dir":"Articles","previous_headings":"3 Reference-based conditional mean imputation - frequentist inference","what":"Impute","title":"rbmi: Inference with Conditional Mean Imputation","text":"can use now function impute() perform imputation original dataset leave-one-samples using results obtained previous step.","code":"references <- c(\"DRUG\" = \"PLACEBO\", \"PLACEBO\" = \"PLACEBO\") imputeObj <- impute(drawObj, references) imputeObj #> #> Imputation Object #> ----------------- #> Number of Imputed Datasets: 1 + 172 #> Fraction of Missing Data (Original Dataset): #> 4: 0% #> 5: 8% #> 6: 13% #> 7: 25% #> References: #> DRUG -> PLACEBO #> PLACEBO -> PLACEBO"},{"path":"/articles/CondMean_Inference.html","id":"analyse","dir":"Articles","previous_headings":"3 Reference-based conditional mean imputation - frequentist inference","what":"Analyse","title":"rbmi: Inference with Conditional Mean Imputation","text":"datasets imputed, can call analyse() function apply complete-data analysis model (ANCOVA) imputed dataset.","code":"# Set analysis variables using rbmi function \"set_vars\" vars_an <- set_vars( group = vars$group, visit = vars$visit, outcome = vars$outcome, covariates = \"BASVAL\" ) # Analyse MAR imputation with derived delta adjustment anaObj <- analyse( imputeObj, rbmi::ancova, vars = vars_an ) anaObj #> #> Analysis Object #> --------------- #> Number of Results: 1 + 172 #> Analysis Function: rbmi::ancova #> Delta Applied: FALSE #> Analysis Estimates: #> trt_4 #> lsm_ref_4 #> lsm_alt_4 #> trt_5 #> lsm_ref_5 #> lsm_alt_5 #> trt_6 #> lsm_ref_6 #> lsm_alt_6 #> trt_7 #> lsm_ref_7 #> lsm_alt_7"},{"path":"/articles/CondMean_Inference.html","id":"pool","dir":"Articles","previous_headings":"3 Reference-based conditional mean imputation - frequentist inference","what":"Pool","title":"rbmi: Inference with Conditional Mean Imputation","text":"Finally, can extract treatment effect estimates perform inference using jackknife variance estimator. done calling pool() function. gives estimated treatment effect 2.13 (95% CI 0.44 3.81) last visit associated p-value 0.013.","code":"poolObj <- pool(anaObj) poolObj #> #> Pool Object #> ----------- #> Number of Results Combined: 1 + 172 #> Method: jackknife #> Confidence Level: 0.95 #> Alternative: two.sided #> #> Results: #> #> ================================================== #> parameter est se lci uci pval #> -------------------------------------------------- #> trt_4 -0.092 0.695 -1.453 1.27 0.895 #> lsm_ref_4 -1.616 0.588 -2.767 -0.464 0.006 #> lsm_alt_4 -1.708 0.396 -2.484 -0.931 <0.001 #> trt_5 1.305 0.878 -0.416 3.027 0.137 #> lsm_ref_5 -4.133 0.688 -5.481 -2.785 <0.001 #> lsm_alt_5 -2.828 0.604 -4.011 -1.645 <0.001 #> trt_6 1.929 0.862 0.239 3.619 0.025 #> lsm_ref_6 -6.088 0.671 -7.402 -4.773 <0.001 #> lsm_alt_6 -4.159 0.686 -5.503 -2.815 <0.001 #> trt_7 2.126 0.858 0.444 3.807 0.013 #> lsm_ref_7 -6.965 0.685 -8.307 -5.622 <0.001 #> lsm_alt_7 -4.839 0.762 -6.333 -3.346 <0.001 #> --------------------------------------------------"},{"path":"/articles/CondMean_Inference.html","id":"reference-based-conditional-mean-imputation---information-anchored-inference","dir":"Articles","previous_headings":"","what":"Reference-based conditional mean imputation - information-anchored inference","title":"rbmi: Inference with Conditional Mean Imputation","text":"section, present estimation process based conditional mean imputation combined jackknife can adapted obtain information-anchored variance following proposal Lu (2021).","code":""},{"path":"/articles/CondMean_Inference.html","id":"draws-1","dir":"Articles","previous_headings":"4 Reference-based conditional mean imputation - information-anchored inference","what":"Draws","title":"rbmi: Inference with Conditional Mean Imputation","text":"code pre-processing dataset “draws” step equivalent code provided frequentist inference. Please refer section details step.","code":"library(rbmi) library(dplyr) dat <- antidepressant_data # Use expand_locf to add rows corresponding to visits with missing outcomes to # the dataset dat <- expand_locf( dat, PATIENT = levels(dat$PATIENT), # expand by PATIENT and VISIT VISIT = levels(dat$VISIT), vars = c(\"BASVAL\", \"THERAPY\"), # fill with LOCF BASVAL and THERAPY group = c(\"PATIENT\"), order = c(\"PATIENT\", \"VISIT\") ) # create data_ice and set the imputation strategy to JR for # each patient with at least one missing observation dat_ice <- dat %>% arrange(PATIENT, VISIT) %>% filter(is.na(CHANGE)) %>% group_by(PATIENT) %>% slice(1) %>% ungroup() %>% select(PATIENT, VISIT) %>% mutate(strategy = \"JR\") # In this dataset, subject 3618 has an intermittent missing values which # does not correspond to a study drug discontinuation. We therefore remove # this subject from `dat_ice`. (In the later imputation step, it will # automatically be imputed under the default MAR assumption.) dat_ice <- dat_ice[-which(dat_ice$PATIENT == 3618),] # Define the names of key variables in our dataset and # the covariates included in the imputation model using `set_vars()` vars <- set_vars( outcome = \"CHANGE\", visit = \"VISIT\", subjid = \"PATIENT\", group = \"THERAPY\", covariates = c(\"BASVAL*VISIT\", \"THERAPY*VISIT\") ) # Define which imputation method to use (here: conditional mean imputation # with jackknife as resampling) method <- method_condmean(type = \"jackknife\") # Create samples for the imputation parameters by running the draws() function drawObj <- draws( data = dat, data_ice = dat_ice, vars = vars, method = method, quiet = TRUE ) drawObj"},{"path":"/articles/CondMean_Inference.html","id":"imputation-step-including-calculation-of-delta-adjustment","dir":"Articles","previous_headings":"4 Reference-based conditional mean imputation - information-anchored inference","what":"Imputation step including calculation of delta-adjustment","title":"rbmi: Inference with Conditional Mean Imputation","text":"proposal Lu (2021) replace reference-based imputation MAR imputation combined delta-adjustment delta selected data-driven way match reference-based estimator. rbmi, implemented first performing imputation defined reference-based imputation strategy (JR) well MAR separately. Second, delta-adjustment defined difference conditional mean imputation reference-based MAR imputation, respectively, original dataset. simplify implementation, written function get_delta_match_refBased performs step. function takes input arguments draws object, data_ice (.e. data.frame containing information intercurrent events imputation strategies), references, named vector identifies references used reference-based imputation methods. function returns list containing imputation objects reference-based MAR imputation, plus data.frame contains delta-adjustment.","code":"#' Get delta adjustment that matches reference-based imputation #' #' @param draws: A `draws` object created by `draws()`. #' @param data_ice: `data.frame` containing the information about the intercurrent #' events and the imputation strategies. Must represent the desired imputation #' strategy and not the MAR-variant. #' @param references: A named vector. Identifies the references to be used #' for reference-based imputation methods. #' #' @return #' The function returns a list containing the imputation objects under both #' reference-based and MAR imputation, plus a `data.frame` which contains the #' delta-adjustment. #' #' @seealso `draws()`, `impute()`. get_delta_match_refBased <- function(draws, data_ice, references) { # Impute according to `data_ice` imputeObj <- impute( draws = drawObj, update_strategy = data_ice, references = references ) vars <- imputeObj$data$vars # Access imputed dataset (index=1 for method_condmean(type = \"jackknife\")) cmi <- extract_imputed_dfs(imputeObj, index = 1, idmap = TRUE)[[1]] idmap <- attributes(cmi)$idmap cmi <- cmi[, c(vars$subjid, vars$visit, vars$outcome)] colnames(cmi)[colnames(cmi) == vars$outcome] <- \"y_imp\" # Map back original patients id since rbmi re-code ids to ensure id uniqueness cmi[[vars$subjid]] <- idmap[match(cmi[[vars$subjid]], names(idmap))] # Derive conditional mean imputations under MAR dat_ice_MAR <- data_ice dat_ice_MAR[[vars$strategy]] <- \"MAR\" # Impute under MAR # Note that in this specific context, it is desirable that an update # from a reference-based strategy to MAR uses the exact same data for # fitting the imputation models, i.e. that available post-ICE data are # omitted from the imputation model for both. This is the case when # using argument update_strategy in function impute(). # However, for other settings (i.e. if one is interested in switching to # a standard MAR imputation strategy altogether), this behavior is # undesirable and, consequently, the function throws a warning which # we suppress here. suppressWarnings( imputeObj_MAR <- impute( draws, update_strategy = dat_ice_MAR ) ) # Access imputed dataset (index=1 for method_condmean(type = \"jackknife\")) cmi_MAR <- extract_imputed_dfs(imputeObj_MAR, index = 1, idmap = TRUE)[[1]] idmap <- attributes(cmi_MAR)$idmap cmi_MAR <- cmi_MAR[, c(vars$subjid, vars$visit, vars$outcome)] colnames(cmi_MAR)[colnames(cmi_MAR) == vars$outcome] <- \"y_MAR\" # Map back original patients id since rbmi re-code ids to ensure id uniqueness cmi_MAR[[vars$subjid]] <- idmap[match(cmi_MAR[[vars$subjid]], names(idmap))] # Derive delta adjustment \"aligned with ref-based imputation\", # i.e. difference between ref-based imputation and MAR imputation delta_adjust <- merge(cmi, cmi_MAR, by = c(vars$subjid, vars$visit), all = TRUE) delta_adjust$delta <- delta_adjust$y_imp - delta_adjust$y_MAR ret_obj <- list( imputeObj = imputeObj, imputeObj_MAR = imputeObj_MAR, delta_adjust = delta_adjust ) return(ret_obj) } references <- c(\"DRUG\" = \"PLACEBO\", \"PLACEBO\" = \"PLACEBO\") res_delta_adjust <- get_delta_match_refBased(drawObj, dat_ice, references)"},{"path":"/articles/CondMean_Inference.html","id":"analyse-1","dir":"Articles","previous_headings":"4 Reference-based conditional mean imputation - information-anchored inference","what":"Analyse","title":"rbmi: Inference with Conditional Mean Imputation","text":"use function analyse() add delta-adjustment perform analysis imputed datasets MAR. analyse() take input argument imputations = res_delta_adjust$imputeObj_MAR, .e. imputation object corresponding MAR imputation (JR imputation). argument delta can used add delta-adjustment prior analysis set delta-adjustment obtained previous step: delta = res_delta_adjust$delta_adjust.","code":"# Set analysis variables using rbmi function \"set_vars\" vars_an <- set_vars( group = vars$group, visit = vars$visit, outcome = vars$outcome, covariates = \"BASVAL\" ) # Analyse MAR imputation with derived delta adjustment anaObj_MAR_delta <- analyse( res_delta_adjust$imputeObj_MAR, rbmi::ancova, delta = res_delta_adjust$delta_adjust, vars = vars_an )"},{"path":"/articles/CondMean_Inference.html","id":"pool-1","dir":"Articles","previous_headings":"4 Reference-based conditional mean imputation - information-anchored inference","what":"Pool","title":"rbmi: Inference with Conditional Mean Imputation","text":"can finally use pool() function extract treatment effect estimate (well estimated marginal means) visit apply jackknife variance estimator analysis estimates imputed leave-one-samples. gives estimated treatment effect 2.13 (95% CI -0.08 4.33) last visit associated p-value 0.058. Per construction delta-adjustment, point estimate identical frequentist analysis. However, standard error much larger (1.12 vs. 0.86). Indeed, information-anchored standard error (resulting inference) similar results Baysesian multiple imputation using Rubin’s rules standard error 1.13 reported quickstart vignette (vignette(topic = \"quickstart\", package = \"rbmi\"). note, shown e.g. Wolbers et al. (2022), hypothesis testing based information-anchored inference conservative, .e. actual type error much lower nominal value. Hence, confidence intervals \\(p\\)-values based information-anchored inference interpreted caution.","code":"poolObj_MAR_delta <- pool(anaObj_MAR_delta) poolObj_MAR_delta #> #> Pool Object #> ----------- #> Number of Results Combined: 1 + 172 #> Method: jackknife #> Confidence Level: 0.95 #> Alternative: two.sided #> #> Results: #> #> ================================================== #> parameter est se lci uci pval #> -------------------------------------------------- #> trt_4 -0.092 0.695 -1.453 1.27 0.895 #> lsm_ref_4 -1.616 0.588 -2.767 -0.464 0.006 #> lsm_alt_4 -1.708 0.396 -2.484 -0.931 <0.001 #> trt_5 1.305 0.944 -0.545 3.156 0.167 #> lsm_ref_5 -4.133 0.738 -5.579 -2.687 <0.001 #> lsm_alt_5 -2.828 0.603 -4.01 -1.646 <0.001 #> trt_6 1.929 0.993 -0.018 3.876 0.052 #> lsm_ref_6 -6.088 0.758 -7.574 -4.602 <0.001 #> lsm_alt_6 -4.159 0.686 -5.504 -2.813 <0.001 #> trt_7 2.126 1.123 -0.076 4.327 0.058 #> lsm_ref_7 -6.965 0.85 -8.63 -5.299 <0.001 #> lsm_alt_7 -4.839 0.763 -6.335 -3.343 <0.001 #> --------------------------------------------------"},{"path":[]},{"path":"/articles/FAQ.html","id":"introduction","dir":"Articles","previous_headings":"","what":"Introduction","title":"rbmi: Frequently Asked Questions","text":"document provides answers common questions rbmi package. intended read rbmi: Quickstart vignette.","code":""},{"path":"/articles/FAQ.html","id":"how-do-the-methods-in-rbmi-compare-to-the-mixed-model-for-repeated-measures-mmrm-implemented-in-the-mmrm-package","dir":"Articles","previous_headings":"1 Introduction","what":"How do the methods in rbmi compare to the mixed model for repeated measures (MMRM) implemented in the mmrm package?","title":"rbmi: Frequently Asked Questions","text":"rbmi designed complement , occasionally, replace standard MMRM analyses clinical trials longitudinal endpoints. Strengths rbmi compared standard MMRM model : rbmi designed allow analyses fully aligned estimand definition. facilitate , implements methods range different missing data assumptions including standard missing--random (MAR), extended MAR (via inclusion time-varying covariates), reference-based missingness, missing--random random (NMAR; via \\(\\delta\\)-adjustments). contrast, standard MMRM model valid standard MAR assumption always plausible. example, standard MAR assumption rather implausible implementing treatment policy strategy intercurrent event “treatment discontinuation” substantial proportion subjects lost--follow-discontinuation. \\(\\delta\\)-adjustment methods implemented rbmi can used sensitivity analyses primary MMRM- rbmi-type analysis. Weaknesses rbmi compared standard MMRM model : MMRM models de-facto standard analysis method decade. rbmi currently less established. rbmi computationally intensive using requires careful planning.","code":""},{"path":"/articles/FAQ.html","id":"how-does-rbmi-compare-to-general-purpose-software-for-multiple-imputation-mi-such-as-mice","dir":"Articles","previous_headings":"1 Introduction","what":"How does rbmi compare to general-purpose software for multiple imputation (MI) such as mice?","title":"rbmi: Frequently Asked Questions","text":"rbmi covers “MMRM-type” settings, .e. settings single longitudinal continuous outcome may missing visits hence require imputation. settings, several advantages general-purpose MI software: rbmi supports imputation range different missing data assumptions whereas general-purpose MI software mostly focused MAR-based imputation. particular, unclear implement jump reference (JR) copy increments reference (CIR) methods software. rbmi interface fully streamlined setting arguably makes implementation straightforward general-purpose MI software. MICE algorithm stochastic inference always based Rubin’s rules. contrast, method “conditional mean imputation plus jackknifing” (method=\"method_condmean(type = \"jackknife\")\") rbmi require tuning parameters, fully deterministic, provides frequentist-consistent inference also reference-based imputations (Rubin’s rule conservative leading actual type error rates can far nominal values). However, rbmi much limited functionality general-purpose MI software.","code":""},{"path":"/articles/FAQ.html","id":"how-to-handle-missing-data-in-baseline-covariates-in-rbmi","dir":"Articles","previous_headings":"1 Introduction","what":"How to handle missing data in baseline covariates in rbmi?","title":"rbmi: Frequently Asked Questions","text":"rbmi support imputation missing baseline covariates. Therefore, missing baseline covariates need handled outside rbmi. best approach handling missing baseline covariates needs made case--case basis context randomized trials, relatively simple approach often sufficient (White Thompson (2005)).","code":""},{"path":"/articles/FAQ.html","id":"why-does-rbmi-by-default-use-an-ancova-analysis-model-and-not-an-mmrm-analysis-model","dir":"Articles","previous_headings":"1 Introduction","what":"Why does rbmi by default use an ANCOVA analysis model and not an MMRM analysis model?","title":"rbmi: Frequently Asked Questions","text":"theoretical justification conditional mean imputation method requires analysis model leads point estimator linear function outcome vector (Wolbers et al. (2022)). case ANCOVA general MMRM models. imputation methods, ANCOVA MMRM valid analysis methods. MMRM analysis model implemented providing custom analysis function analyse() function. expalanations, also cite end section 2.4 conditional mean imputation paper (Wolbers et al. (2022)): proof relies fact ANCOVA estimator linear function outcome vector. complete data, ANCOVA estimator leads identical parameter estimates MMRM model longitudinal outcomes arbitrary common covariance structure across treatment groups treatment--visit interactions well covariate--visit-interactions included analysis model covariates,17 (p. 197). Hence, proof also applies MMRM models. expect conditional mean imputation also valid general MMRM model used analysis involved argument required formally justify .","code":""},{"path":"/articles/FAQ.html","id":"how-can-i-analyse-the-change-from-baseline-in-the-analysis-model-when-imputation-was-done-on-the-original-outcomes","dir":"Articles","previous_headings":"1 Introduction","what":"How can I analyse the change-from-baseline in the analysis model when imputation was done on the original outcomes?","title":"rbmi: Frequently Asked Questions","text":"can achieved using custom analysis functions outlined Section 7 Advanced Vignette. e.g.","code":"ancova_modified <- function(data, ...) { data2 <- data %>% mutate(ENDPOINT = ENDPOINT - BASELINE) rbmi::ancova(data2, ...) } anaObj <- rbmi::analyse( imputeObj, ancova_modified, vars = vars )"},{"path":"/articles/advanced.html","id":"introduction","dir":"Articles","previous_headings":"","what":"Introduction","title":"rbmi: Advanced Functionality","text":"purpose vignette provide overview advanced features rbmi package. sections vignette relatively self-contained, .e. readers able jump directly section covers functionality interested .","code":""},{"path":"/articles/advanced.html","id":"sec:dataSimul","dir":"Articles","previous_headings":"","what":"Data simulation using function simulate_data()","title":"rbmi: Advanced Functionality","text":"order demonstrate advanced functions first create simulated dataset rbmi function simulate_data(). simulate_data() function generates data randomized clinical trial longitudinal continuous outcomes two different types intercurrent events (ICEs). One intercurrent event (ICE1) may thought discontinuation study treatment due study drug condition related (SDCR) reasons. event (ICE2) may thought discontinuation study treatment due study drug condition related (NSDCR) reasons. purpose vignette, simulate data similarly simulation study reported Wolbers et al. (2022) (though change simulation parameters) include one ICE type (ICE1). Specifically, simulate 1:1 randomized trial active drug (intervention) versus placebo (control) 100 subjects per group 6 post-baseline assessments (bi-monthly visits 12 months) following assumptions: mean outcome trajectory placebo group increases linearly 50 baseline (visit 0) 60 visit 6, .e. slope 10 points/year. mean outcome trajectory intervention group identical placebo group visit 2. visit 2 onward, slope decreases 50% 5 points/year. covariance structure baseline follow-values groups implied random intercept slope model standard deviation 5 intercept slope, correlation 0.25. addition, independent residual error standard deviation 2.5 added assessment. probability study drug discontinuation visit calculated according logistic model depends observed outcome visit. Specifically, visit-wise discontinuation probability 2% 3% control intervention group, respectively, specified case observed outcome equal 50 (mean value baseline). odds discontinuation simulated increase +10% +1 point increase observed outcome. Study drug discontinuation simulated effect mean trajectory placebo group. intervention group, subjects discontinue follow slope mean trajectory placebo group time point onward. compatible copy increments reference (CIR) assumption. Study drop-study drug discontinuation visit occurs probability 50% leading missing outcome data time point onward. function simulate_data() requires 3 arguments (see function documentation help(simulate_data) details): pars_c: simulation parameters control group pars_t: simulation parameters intervention group post_ice1_traj: Specifies observed outcomes ICE1 simulated , report data according specifications can simulated function simulate_data():","code":"library(rbmi) library(dplyr) library(ggplot2) library(purrr) set.seed(122) n <- 100 time <- c(0, 2, 4, 6, 8, 10, 12) # Mean trajectory control muC <- c(50.0, 51.66667, 53.33333, 55.0, 56.66667, 58.33333, 60.0) # Mean trajectory intervention muT <- c(50.0, 51.66667, 53.33333, 54.16667, 55.0, 55.83333, 56.66667) # Create Sigma sd_error <- 2.5 covRE <- rbind( c(25.0, 6.25), c(6.25, 25.0) ) Sigma <- cbind(1, time / 12) %*% covRE %*% rbind(1, time / 12) + diag(sd_error^2, nrow = length(time)) # Set probability of discontinuation probDisc_C <- 0.02 probDisc_T <- 0.03 or_outcome <- 1.10 # +1 point increase => +10% odds of discontinuation # Set drop-out rate following discontinuation prob_dropout <- 0.5 # Set simulation parameters of the control group parsC <- set_simul_pars( mu = muC, sigma = Sigma, n = n, prob_ice1 = probDisc_C, or_outcome_ice1 = or_outcome, prob_post_ice1_dropout = prob_dropout ) # Set simulation parameters of the intervention group parsT <- parsC parsT$mu <- muT parsT$prob_ice1 <- probDisc_T # Set assumption about post-ice trajectory post_ice_traj <- \"CIR\" # Simulate data data <- simulate_data( pars_c = parsC, pars_t = parsT, post_ice1_traj = post_ice_traj ) head(data) #> id visit group outcome_bl outcome_noICE ind_ice1 ind_ice2 dropout_ice1 #> 1 id_1 0 Control 57.32704 57.32704 0 0 0 #> 2 id_1 1 Control 57.32704 54.69751 1 0 1 #> 3 id_1 2 Control 57.32704 58.60702 1 0 1 #> 4 id_1 3 Control 57.32704 61.50119 1 0 1 #> 5 id_1 4 Control 57.32704 56.68363 1 0 1 #> 6 id_1 5 Control 57.32704 66.14799 1 0 1 #> outcome #> 1 57.32704 #> 2 NA #> 3 NA #> 4 NA #> 5 NA #> 6 NA # As a simple descriptive of the simulated data, summarize the number of subjects with ICEs and missing data data %>% group_by(id) %>% summarise( group = group[1], any_ICE = (any(ind_ice1 == 1)), any_NA = any(is.na(outcome))) %>% group_by(group) %>% summarise( subjects_with_ICE = sum(any_ICE), subjects_with_missings = sum(any_NA) ) #> # A tibble: 2 × 3 #> group subjects_with_ICE subjects_with_missings #> #> 1 Control 18 8 #> 2 Intervention 25 14"},{"path":"/articles/advanced.html","id":"sec:postICEobs","dir":"Articles","previous_headings":"","what":"Handling of observed post-ICE data in rbmi under reference-based imputation","title":"rbmi: Advanced Functionality","text":"rbmi always uses non-missing outcome data input data set, .e. data never overwritten imputation step removed analysis step. implies data considered irrelevant treatment effect estimation (e.g. data ICE estimand specified hypothetical strategy), data need removed input data set user prior calling rbmi functions. imputation missing random (MAR) strategy, observed outcome data also included fitting base imputation model. However, ICEs handled using reference-based imputation methods (CIR, CR, JR), rbmi excludes observed post-ICE data base imputation model. data excluded, base imputation model mistakenly estimate mean trajectories based mixture observed pre- post-ICE data relevant reference-based imputations. However, observed post-ICE data added back data set fitting base imputation model included subsequent imputation analysis steps. Post-ICE data control reference group also excluded base imputation model user specifies reference-based imputation strategy ICEs. ensures ICE impact data included base imputation model regardless whether ICE occurred control intervention group. hand, imputation reference group based MAR assumption even reference-based imputation methods may preferable settings include post-ICE data control group base imputation model. can implemented specifying MAR strategy ICE control group reference-based strategy ICE intervention group. use latter approach example . simulated trial data section 2 assumed outcomes intervention group observed ICE “treatment discontinuation” follow increments observed control group. Thus imputation missing data intervention group treatment discontinuation might performed reference-based copy increments reference (CIR) assumption. Specifically, implement estimator following assumptions: endpoint interest change outcome baseline visit. imputation model includes treatment group, (categorical) visit, treatment--visit interactions, baseline outcome, baseline outcome--visit interactions covariates. imputation model assumes common unstructured covariance matrix treatment groups control group, missing data imputed MAR whereas intervention group, missing post-ICE data imputed CIR assumption analysis model endpoint imputed datasets separate ANCOVA model visit treatment group primary covariate adjustment baseline outcome value. illustration purposes, chose MI based approximate Bayesian posterior draws 20 random imputations demanding computational perspective. practical applications, number random imputations may need increased. Moreover, imputations also supported rbmi. guidance regarding choice imputation approach, refer user comparison implemented approaches Section 3.9 “Statistical Specifications” vignette (vignette(\"stat_specs\", package = \"rbmi\")). first report code set variables imputation analysis models. yet familiar syntax, recommend first check “quickstart” vignette (vignette(\"quickstart\", package = \"rbmi\")). chosen imputation method can set function method_approxbayes() follows: can now sequentially call 4 key functions rbmi perform multiple imputation. Please note management observed post-ICE data performed without additional complexity user. draws() automatically excludes post-ICE data handled reference-based method (keeps post-ICE data handled using MAR) using information provided argument data_ice. impute() impute truly missing data data[[vars$outcome]]. last output gives estimated difference -4.537 (95% CI -6.420 -2.655) two groups last visit associated p-value lower 0.001.","code":"# Create data_ice including the subject's first visit affected by the ICE and the imputation strategy # Imputation strategy for post-ICE data is CIR in the intervention group and MAR for the control group # (note that ICEs which are handled using MAR are optional and do not impact the analysis # because imputation of missing data under MAR is the default) data_ice_CIR <- data %>% group_by(id) %>% filter(ind_ice1 == 1) %>% # select visits with ICEs mutate(strategy = ifelse(group == \"Intervention\", \"CIR\", \"MAR\")) %>% summarise( visit = visit[1], # Select first visit affected by the ICE strategy = strategy[1] ) # Compute endpoint of interest: change from baseline and # remove rows corresponding to baseline visits data <- data %>% filter(visit != 0) %>% mutate( change = outcome - outcome_bl, visit = factor(visit, levels = unique(visit)) ) # Define key variables for the imputation and analysis models vars <- set_vars( subjid = \"id\", visit = \"visit\", outcome = \"change\", group = \"group\", covariates = c(\"visit*outcome_bl\", \"visit*group\"), strategy = \"strategy\" ) vars_an <- vars vars_an$covariates <- \"outcome_bl\" method <- method_approxbayes(n_sample = 20) draw_obj <- draws( data = data, data_ice = data_ice_CIR, vars = vars, method = method, quiet = TRUE, ncores = 2 ) impute_obj_CIR <- impute( draw_obj, references = c(\"Control\" = \"Control\", \"Intervention\" = \"Control\") ) ana_obj_CIR <- analyse( impute_obj_CIR, vars = vars_an ) pool_obj_CIR <- pool(ana_obj_CIR) pool_obj_CIR #> #> Pool Object #> ----------- #> Number of Results Combined: 20 #> Method: rubin #> Confidence Level: 0.95 #> Alternative: two.sided #> #> Results: #> #> ================================================== #> parameter est se lci uci pval #> -------------------------------------------------- #> trt_1 -0.486 0.512 -1.496 0.524 0.343 #> lsm_ref_1 2.62 0.362 1.907 3.333 <0.001 #> lsm_alt_1 2.133 0.362 1.42 2.847 <0.001 #> trt_2 -0.066 0.542 -1.135 1.004 0.904 #> lsm_ref_2 3.707 0.384 2.95 4.464 <0.001 #> lsm_alt_2 3.641 0.383 2.885 4.397 <0.001 #> trt_3 -1.782 0.607 -2.979 -0.585 0.004 #> lsm_ref_3 5.841 0.428 4.997 6.685 <0.001 #> lsm_alt_3 4.059 0.428 3.214 4.904 <0.001 #> trt_4 -2.518 0.692 -3.884 -1.152 <0.001 #> lsm_ref_4 7.656 0.492 6.685 8.627 <0.001 #> lsm_alt_4 5.138 0.488 4.176 6.1 <0.001 #> trt_5 -3.658 0.856 -5.346 -1.97 <0.001 #> lsm_ref_5 9.558 0.598 8.379 10.737 <0.001 #> lsm_alt_5 5.9 0.608 4.699 7.101 <0.001 #> trt_6 -4.537 0.954 -6.42 -2.655 <0.001 #> lsm_ref_6 11.048 0.666 9.735 12.362 <0.001 #> lsm_alt_6 6.511 0.674 5.181 7.841 <0.001 #> --------------------------------------------------"},{"path":"/articles/advanced.html","id":"efficiently-changing-reference-based-imputation-strategies","dir":"Articles","previous_headings":"","what":"Efficiently changing reference-based imputation strategies","title":"rbmi: Advanced Functionality","text":"draws() function far computationally intensive function rbmi. settings, may important explore impact change reference-based imputation strategy results. change affect imputation model affect subsequent imputation step. order allow changes imputation strategy without re-run draws() function, function impute() additional argument update_strategies. However, please note functionality comes important limitations: described beginning Section 3, post-ICE outcomes included input dataset base imputation model imputation method MAR excluded reference-based imputation methods (CIR, CR, JR). Therefore, updata_strategies applied imputation strategy changed MAR non-MAR strategy presence observed post-ICE outcomes. Similarly, change non-MAR strategy MAR triggers warning presence observed post-ICE outcomes base imputation model fitted relevant data MAR. Finally, update_strategies applied timing ICEs changed (argument data_ice) addition imputation strategy. example, described analysis copy increments reference (CIR) assumption previous section. Let’s assume want change strategy jump reference imputation strategy sensitivity analysis. can efficiently implemented using update_strategies follows: imputations jump reference assumption, get estimated difference -4.360 (95% CI -6.238 -2.482) two groups last visit associated p-value <0.001.","code":"# Change ICE strategy from CIR to JR data_ice_JR <- data_ice_CIR %>% mutate(strategy = ifelse(strategy == \"CIR\", \"JR\", strategy)) impute_obj_JR <- impute( draw_obj, references = c(\"Control\" = \"Control\", \"Intervention\" = \"Control\"), update_strategy = data_ice_JR ) ana_obj_JR <- analyse( impute_obj_JR, vars = vars_an ) pool_obj_JR <- pool(ana_obj_JR) pool_obj_JR #> #> Pool Object #> ----------- #> Number of Results Combined: 20 #> Method: rubin #> Confidence Level: 0.95 #> Alternative: two.sided #> #> Results: #> #> ================================================== #> parameter est se lci uci pval #> -------------------------------------------------- #> trt_1 -0.485 0.513 -1.496 0.526 0.346 #> lsm_ref_1 2.609 0.363 1.892 3.325 <0.001 #> lsm_alt_1 2.124 0.361 1.412 2.836 <0.001 #> trt_2 -0.06 0.535 -1.115 0.995 0.911 #> lsm_ref_2 3.694 0.378 2.948 4.441 <0.001 #> lsm_alt_2 3.634 0.381 2.882 4.387 <0.001 #> trt_3 -1.767 0.598 -2.948 -0.587 0.004 #> lsm_ref_3 5.845 0.422 5.012 6.677 <0.001 #> lsm_alt_3 4.077 0.432 3.225 4.93 <0.001 #> trt_4 -2.529 0.686 -3.883 -1.175 <0.001 #> lsm_ref_4 7.637 0.495 6.659 8.614 <0.001 #> lsm_alt_4 5.108 0.492 4.138 6.078 <0.001 #> trt_5 -3.523 0.856 -5.212 -1.833 <0.001 #> lsm_ref_5 9.554 0.61 8.351 10.758 <0.001 #> lsm_alt_5 6.032 0.611 4.827 7.237 <0.001 #> trt_6 -4.36 0.952 -6.238 -2.482 <0.001 #> lsm_ref_6 11.003 0.676 9.669 12.337 <0.001 #> lsm_alt_6 6.643 0.687 5.287 8 <0.001 #> --------------------------------------------------"},{"path":"/articles/advanced.html","id":"imputation-under-mar-with-time-varying-covariates","dir":"Articles","previous_headings":"","what":"Imputation under MAR with time-varying covariates","title":"rbmi: Advanced Functionality","text":"Guizzaro et al. (2021) suggested implement treatment policy strategy via imputation MAR assumption conditioning subject’s ICE status, .e. impute missing post-ICE data based observed post-ICE data. One possible implementation proposal add time-varying covariates imputation model. case study implements proposal compares reference-based imputation methods estimators early Parkinson’s disease can found Noci et al. (2021). settings, may carried including binary time-varying indicator subject’s ICE status visit (defined 0 pre-ICE visits 1 post-ICE visits) imputation model. However, simulated data introduced section 2, may plausible assume treatment discontinuation leads change “slope” mean outcome trajectory. can implemented including time-varying covariate equal 0 visits prior treatment discontinuation equal time treatment discontinuation subsequent visits. regression coefficient corresponding change post-ICE “slope” allowed depend assigned treatment group, .e. imputation model include interaction time-varying covariate treatment group. Let’s first define time-varying covariate: can include time-varying covariate imputation model, crossed group variable: now sequentially call 4 key rbmi functions:","code":"data <- data %>% group_by(id) %>% mutate(time_from_ice1 = cumsum(ind_ice1)*2/12 ) # multiplication by 2/12 because visits are bi-monthly vars_tv <- set_vars( subjid = \"id\", visit = \"visit\", outcome = \"change\", group = \"group\", covariates = c(\"visit*outcome_bl\", \"visit*group\", \"time_from_ice1*group\"), strategy = \"strategy\" ) draw_obj <- draws( data = data, data_ice = NULL, # if NULL, MAR is assumed for all missing data vars = vars_tv, method = method, quiet = TRUE ) impute_obj_tv <- impute( draw_obj, references = c(\"Control\" = \"Control\", \"Intervention\" = \"Intervention\") ) ana_obj_tv <- analyse( impute_obj_tv, vars = vars_an ) pool(ana_obj_tv) #> #> Pool Object #> ----------- #> Number of Results Combined: 20 #> Method: rubin #> Confidence Level: 0.95 #> Alternative: two.sided #> #> Results: #> #> ================================================== #> parameter est se lci uci pval #> -------------------------------------------------- #> trt_1 -0.492 0.515 -1.507 0.524 0.341 #> lsm_ref_1 2.623 0.362 1.908 3.338 <0.001 #> lsm_alt_1 2.131 0.366 1.409 2.854 <0.001 #> trt_2 0.018 0.55 -1.067 1.103 0.974 #> lsm_ref_2 3.697 0.382 2.943 4.45 <0.001 #> lsm_alt_2 3.715 0.394 2.936 4.493 <0.001 #> trt_3 -1.802 0.614 -3.015 -0.59 0.004 #> lsm_ref_3 5.815 0.429 4.97 6.661 <0.001 #> lsm_alt_3 4.013 0.441 3.142 4.884 <0.001 #> trt_4 -2.543 0.704 -3.932 -1.154 <0.001 #> lsm_ref_4 7.609 0.486 6.65 8.568 <0.001 #> lsm_alt_4 5.066 0.516 4.046 6.086 <0.001 #> trt_5 -3.739 0.879 -5.475 -2.004 <0.001 #> lsm_ref_5 9.499 0.606 8.302 10.695 <0.001 #> lsm_alt_5 5.759 0.636 4.502 7.017 <0.001 #> trt_6 -4.685 0.98 -6.622 -2.748 <0.001 #> lsm_ref_6 10.988 0.667 9.67 12.305 <0.001 #> lsm_alt_6 6.302 0.712 4.894 7.711 <0.001 #> --------------------------------------------------"},{"path":"/articles/advanced.html","id":"custom-imputation-strategies","dir":"Articles","previous_headings":"","what":"Custom imputation strategies","title":"rbmi: Advanced Functionality","text":"following imputation strategies implemented rbmi: Missing Random (MAR) Jump Reference (JR) Copy Reference (CR) Copy Increments Reference (CIR) Last Mean Carried Forward (LMCF) addition, rbmi allows user implement imputation strategy. , user needs three things: Define function implementing new imputation strategy. Specify patients use strategy data_ice dataset provided draws(). Provide imputation strategy function impute(). imputation strategy function must take 3 arguments (pars_group, pars_ref, index_mar) calculates mean covariance matrix subject’s marginal imputation distribution applied subjects strategy applies. , pars_group contains predicted mean trajectory (pars_group$mu, numeric vector) covariance matrix (pars_group$sigma) subject conditional assigned treatment group covariates. pars_ref contains corresponding mean trajectory covariance matrix conditional reference group subject’s covariates. index_mar logical vector specifies visit whether visit unaffected ICE handled using non-MAR method . example, user can check CIR strategy implemented looking function strategy_CIR(). illustrate simple example, assume new strategy implemented follows: - marginal mean imputation distribution equal marginal mean trajectory subject according assigned group covariates ICE. - ICE marginal mean imputation distribution equal average visit-wise marginal means based subjects covariates assigned group reference group, respectively. - covariance matrix marginal imputation distribution, covariance matrix assigned group taken. , first need define imputation function example coded follows: example showing use: incorporate rbmi, data_ice needs updated strategy AVG specified visits affected ICE. Additionally, function needs provided impute() via getStrategies() function shown : , analysis proceed calling analyse() pool() .","code":"strategy_CIR #> function (pars_group, pars_ref, index_mar) #> { #> if (all(index_mar)) { #> return(pars_group) #> } #> else if (all(!index_mar)) { #> return(pars_ref) #> } #> mu <- pars_group$mu #> last_mar <- which(!index_mar)[1] - 1 #> increments_from_last_mar_ref <- pars_ref$mu[!index_mar] - #> pars_ref$mu[last_mar] #> mu[!index_mar] <- mu[last_mar] + increments_from_last_mar_ref #> sigma <- compute_sigma(sigma_group = pars_group$sigma, sigma_ref = pars_ref$sigma, #> index_mar = index_mar) #> pars <- list(mu = mu, sigma = sigma) #> return(pars) #> } #> #> strategy_AVG <- function(pars_group, pars_ref, index_mar) { mu_mean <- (pars_group$mu + pars_ref$mu) / 2 x <- pars_group x$mu[!index_mar] <- mu_mean[!index_mar] return(x) } pars_group <- list( mu = c(1, 2, 3), sigma = as_vcov(c(1, 3, 2), c(0.4, 0.5, 0.45)) ) pars_ref <- list( mu = c(5, 6, 7), sigma = as_vcov(c(2, 1, 1), c(0.7, 0.8, 0.5)) ) index_mar <- c(TRUE, TRUE, FALSE) strategy_AVG(pars_group, pars_ref, index_mar) #> $mu #> [1] 1 2 5 #> #> $sigma #> [,1] [,2] [,3] #> [1,] 1.0 1.2 1.0 #> [2,] 1.2 9.0 2.7 #> [3,] 1.0 2.7 4.0 data_ice_AVG <- data_ice_CIR %>% mutate(strategy = ifelse(strategy == \"CIR\", \"AVG\", strategy)) draw_obj <- draws( data = data, data_ice = data_ice_AVG, vars = vars, method = method, quiet = TRUE ) impute_obj <- impute( draw_obj, references = c(\"Control\" = \"Control\", \"Intervention\" = \"Control\"), strategies = getStrategies(AVG = strategy_AVG) )"},{"path":"/articles/advanced.html","id":"custom-analysis-functions","dir":"Articles","previous_headings":"","what":"Custom analysis functions","title":"rbmi: Advanced Functionality","text":"default rbmi analyse data using ancova() function. analysis function fits ANCOVA model outcomes visit separately, returns “treatment effect” estimate well corresponding least square means group. user wants perform different analysis, return different statistics analysis, can done using custom analysis function. Beware validity conditional mean imputation method formally established analysis functions corresponding linear models (ANCOVA) caution required applying alternative analysis functions method. custom analysis function must take data.frame first argument return named list element list containing minimum point estimate, called est. method method_bayes() method_approxbayes(), list must additionally contain standard error (element se) , available, degrees freedom complete-data analysis model (element df). simple example, replicate ANCOVA analysis last visit CIR-based imputations user-defined analysis function : second example, assume supplementary analysis user wants compare proportion subjects change baseline >10 points last visit treatment groups baseline outcome additional covariate. lead following basic analysis function: Note user wants rbmi use normal approximation pooled test statistics, degrees freedom need set df = NA (per example). degrees freedom complete data test statistics known degrees freedom set df = Inf, rbmi pools degrees freedom across imputed datasets according rule Barnard Rubin (see “Statistical Specifications” vignette (vignette(\"stat_specs\", package = \"rbmi\") details). According rule, infinite degrees freedom complete data analysis imply pooled degrees freedom also infinite. Rather, case pooled degrees freedom (M-1)/lambda^2, M number imputations lambda fraction missing information (see Barnard Rubin (1999) details).","code":"compare_change_lastvisit <- function(data, ...) { fit <- lm(change ~ group + outcome_bl, data = data, subset = (visit == 6) ) res <- list( trt = list( est = coef(fit)[\"groupIntervention\"], se = sqrt(vcov(fit)[\"groupIntervention\", \"groupIntervention\"]), df = df.residual(fit) ) ) return(res) } ana_obj_CIR6 <- analyse( impute_obj_CIR, fun = compare_change_lastvisit, vars = vars_an ) pool(ana_obj_CIR6) #> #> Pool Object #> ----------- #> Number of Results Combined: 20 #> Method: rubin #> Confidence Level: 0.95 #> Alternative: two.sided #> #> Results: #> #> ================================================= #> parameter est se lci uci pval #> ------------------------------------------------- #> trt -4.537 0.954 -6.42 -2.655 <0.001 #> ------------------------------------------------- compare_prop_lastvisit <- function(data, ...) { fit <- glm( I(change > 10) ~ group + outcome_bl, family = binomial(), data = data, subset = (visit == 6) ) res <- list( trt = list( est = coef(fit)[\"groupIntervention\"], se = sqrt(vcov(fit)[\"groupIntervention\", \"groupIntervention\"]), df = NA ) ) return(res) } ana_obj_prop <- analyse( impute_obj_CIR, fun = compare_prop_lastvisit, vars = vars_an ) pool_obj_prop <- pool(ana_obj_prop) pool_obj_prop #> #> Pool Object #> ----------- #> Number of Results Combined: 20 #> Method: rubin #> Confidence Level: 0.95 #> Alternative: two.sided #> #> Results: #> #> ================================================= #> parameter est se lci uci pval #> ------------------------------------------------- #> trt -1.052 0.314 -1.667 -0.438 0.001 #> ------------------------------------------------- tmp <- as.data.frame(pool_obj_prop) %>% mutate( OR = exp(est), OR.lci = exp(lci), OR.uci = exp(uci) ) %>% select(parameter, OR, OR.lci, OR.uci) tmp #> parameter OR OR.lci OR.uci #> 1 trt 0.3491078 0.188807 0.6455073"},{"path":"/articles/advanced.html","id":"sensitivity-analyses-delta-adjustments-and-tipping-point-analyses","dir":"Articles","previous_headings":"","what":"Sensitivity analyses: Delta adjustments and tipping point analyses","title":"rbmi: Advanced Functionality","text":"Delta-adjustments used impute missing data missing random (NMAR) assumption. reflects belief unobserved outcomes systematically “worse” (“better”) “comparable” observed outcomes. extensive discussion delta-adjustment methods, refer Cro et al. (2020). rbmi, marginal delta-adjustment approach implemented. means delta-adjustment applied dataset data imputation MAR reference-based missing data assumptions prior analysis imputed data. Sensitivity analysis using delta-adjustments can therefore performed without re-fit imputation model. rbmi, implemented via delta argument analyse() function.","code":""},{"path":"/articles/advanced.html","id":"simple-delta-adjustments-and-tipping-point-analyses","dir":"Articles","previous_headings":"8 Sensitivity analyses: Delta adjustments and tipping point analyses","what":"Simple delta adjustments and tipping point analyses","title":"rbmi: Advanced Functionality","text":"delta argument analyse() allows users modify outcome variable prior analysis. , user needs provide data.frame contains columns subject visit (identify observation adjusted) plus additional column called delta specifies value added outcomes prior analysis. delta_template() function supports user creating data.frame: creates skeleton data.frame containing one row per subject visit value delta set 0 observations: Note output delta_template() contains additional information can used properly re-set variable delta. example, assume user wants implement delta-adjustment imputed values CIR described section 3. Specifically, assume fixed “worsening adjustment” +5 points applied imputed values regardless treatment group. programmed follows: approach can used implement tipping point analysis. , apply different delta-adjustments imputed data control intervention group, respectively. Assume delta-adjustments less -5 points +15 points considered implausible clinical perspective. Therefore, vary delta-values group -5 +15 points investigate delta combinations lead “tipping” primary analysis result, defined analysis p-value \\(\\geq 0.05\\). According analysis, significant test result primary analysis CIR tipped non-significant result rather extreme delta-adjustments. Please note real analysis recommended use smaller step size grid used .","code":"dat_delta <- delta_template(imputations = impute_obj_CIR) head(dat_delta) #> id visit group is_mar is_missing is_post_ice strategy delta #> 1 id_1 1 Control TRUE TRUE TRUE MAR 0 #> 2 id_1 2 Control TRUE TRUE TRUE MAR 0 #> 3 id_1 3 Control TRUE TRUE TRUE MAR 0 #> 4 id_1 4 Control TRUE TRUE TRUE MAR 0 #> 5 id_1 5 Control TRUE TRUE TRUE MAR 0 #> 6 id_1 6 Control TRUE TRUE TRUE MAR 0 # Set delta-value to 5 for all imputed (previously missing) outcomes and 0 for all other outcomes dat_delta <- delta_template(imputations = impute_obj_CIR) %>% mutate(delta = is_missing * 5) # Repeat the analyses with the delta-adjusted values and pool results ana_delta <- analyse( impute_obj_CIR, delta = dat_delta, vars = vars_an ) pool(ana_delta) #> #> Pool Object #> ----------- #> Number of Results Combined: 20 #> Method: rubin #> Confidence Level: 0.95 #> Alternative: two.sided #> #> Results: #> #> ================================================== #> parameter est se lci uci pval #> -------------------------------------------------- #> trt_1 -0.482 0.524 -1.516 0.552 0.359 #> lsm_ref_1 2.718 0.37 1.987 3.448 <0.001 #> lsm_alt_1 2.235 0.37 1.505 2.966 <0.001 #> trt_2 -0.016 0.56 -1.12 1.089 0.978 #> lsm_ref_2 3.907 0.396 3.125 4.688 <0.001 #> lsm_alt_2 3.891 0.395 3.111 4.671 <0.001 #> trt_3 -1.684 0.641 -2.948 -0.42 0.009 #> lsm_ref_3 6.092 0.452 5.201 6.983 <0.001 #> lsm_alt_3 4.408 0.452 3.515 5.3 <0.001 #> trt_4 -2.359 0.741 -3.821 -0.897 0.002 #> lsm_ref_4 7.951 0.526 6.913 8.99 <0.001 #> lsm_alt_4 5.593 0.522 4.563 6.623 <0.001 #> trt_5 -3.34 0.919 -5.153 -1.526 <0.001 #> lsm_ref_5 9.899 0.643 8.631 11.168 <0.001 #> lsm_alt_5 6.559 0.653 5.271 7.848 <0.001 #> trt_6 -4.21 1.026 -6.236 -2.184 <0.001 #> lsm_ref_6 11.435 0.718 10.019 12.851 <0.001 #> lsm_alt_6 7.225 0.725 5.793 8.656 <0.001 #> -------------------------------------------------- perform_tipp_analysis <- function(delta_control, delta_intervention) { # Derive delta offset based on control and intervention specific deltas delta_df <- delta_df_init %>% mutate( delta_ctl = (group == \"Control\") * is_missing * delta_control, delta_int = (group == \"Intervention\") * is_missing * delta_intervention, delta = delta_ctl + delta_int ) ana_delta <- analyse( impute_obj_CIR, fun = compare_change_lastvisit, vars = vars_an, delta = delta_df, ) pool_delta <- as.data.frame(pool(ana_delta)) list( trt_effect_6 = pool_delta[[\"est\"]], pval_6 = pool_delta[[\"pval\"]] ) } # Get initial delta template delta_df_init <- delta_template(impute_obj_CIR) tipp_frame_grid <- expand.grid( delta_control = seq(-5, 15, by = 2), delta_intervention = seq(-5, 15, by = 2) ) %>% as_tibble() tipp_frame <- tipp_frame_grid %>% mutate( results_list = map2(delta_control, delta_intervention, perform_tipp_analysis), trt_effect_6 = map_dbl(results_list, \"trt_effect_6\"), pval_6 = map_dbl(results_list, \"pval_6\") ) %>% select(-results_list) %>% mutate( pval = cut( pval_6, c(0, 0.001, 0.01, 0.05, 0.2, 1), right = FALSE, labels = c(\"<0.001\", \"0.001 - <0.01\", \"0.01- <0.05\", \"0.05 - <0.20\", \">= 0.20\") ) ) # Show delta values which lead to non-significant analysis results tipp_frame %>% filter(pval_6 >= 0.05) #> # A tibble: 3 × 5 #> delta_control delta_intervention trt_effect_6 pval_6 pval #> #> 1 -5 15 -1.99 0.0935 0.05 - <0.20 #> 2 -3 15 -2.15 0.0704 0.05 - <0.20 #> 3 -1 15 -2.31 0.0527 0.05 - <0.20 ggplot(tipp_frame, aes(delta_control, delta_intervention, fill = pval)) + geom_raster() + scale_fill_manual(values = c(\"darkgreen\", \"lightgreen\", \"lightyellow\", \"orange\", \"red\"))"},{"path":"/articles/advanced.html","id":"more-flexible-delta-adjustments-using-the-dlag-and-delta-arguments-of-delta_template","dir":"Articles","previous_headings":"8 Sensitivity analyses: Delta adjustments and tipping point analyses","what":"More flexible delta-adjustments using the dlag and delta arguments of delta_template()","title":"rbmi: Advanced Functionality","text":"far, discussed simple delta arguments add value imputed values. However, user may want apply flexible delta-adjustments missing values intercurrent event (ICE) vary magnitude delta adjustment depending far away visit question ICE visit. facilitate creation flexible delta-adjustments, delta_template() function two optional additional arguments delta dlag. delta argument specifies default amount delta applied post-ICE visit, whilst dlag specifies scaling coefficient applied based upon visits proximity first visit affected ICE. default, delta added unobserved (.e. imputed) post-ICE outcomes can changed setting optional argument missing_only = FALSE. usage delta dlag arguments best illustrated examples: Assume setting 4 visits user specified delta = c(5,6,7,8) dlag=c(1,2,3,4). subject first visit affected ICE visit 2, values delta dlag imply following delta offset: , subject delta offset 0 applied visit v1, 6 visit v2, 20 visit v3 44 visit v4. Assume instead, subject’s first visit affected ICE visit 3. , values delta dlag imply following delta offset: apply constant delta value +5 visits affected ICE regardless proximity first ICE visit, one set delta = c(5,5,5,5) dlag = c(1,0,0,0). Alternatively, may straightforward setting call delta_template() function without delta dlag arguments overwrite delta column resulting data.frame described previous section (additionally relying is_post_ice variable). Another way using arguments set delta difference time visits dlag amount delta per unit time. example, let’s say visits occur weeks 1, 5, 6 9 want delta 3 applied week ICE. simplicity, assume ICE occurs immediately subject’s last visit affected ICE. achieved setting delta = c(1,4,1,3) (difference weeks visit) dlag = c(3, 3, 3, 3). Assume subject’s first visit affected ICE visit v2, values delta dlag imply following delta offsets: wrap , show action simulated dataset section 2 imputed datasets based CIR assumption section 3. simulation setting specified follow-visits months 2, 4, 6, 8, 10, 12. Assume want apply delta-adjustment 1 every month ICE unobserved post-ICE visits intervention group . (E.g. ICE occurred immediately month 4 visit, total delta applied missing value month 10 visit 6.) program , first use delta dlag arguments delta_template() set corresponding template data.frame: Next, can use additional metadata variables provided delta_template() manually reset delta values control group back 0: Finally, can use delta data.frame apply desired delta offset analysis:","code":"v1 v2 v3 v4 -------------- 5 6 7 8 # delta assigned to each visit 0 1 2 3 # scaling starting from the first visit after the subjects ICE -------------- 0 6 14 24 # delta * scaling -------------- 0 6 20 44 # cumulative sum (i.e. delta) to be applied to each visit v1 v2 v3 v4 -------------- 5 6 7 8 # delta assigned to each visit 0 0 1 2 # scaling starting from the first visit after the subjects ICE -------------- 0 0 7 16 # delta * scaling -------------- 0 0 7 23 # cumulative sum (i.e. delta) to be applied to each visit v1 v2 v3 v4 -------------- 1 4 1 3 # delta assigned to each visit 0 3 3 3 # scaling starting from the first visit after the subjects ICE -------------- 0 12 3 9 # delta * scaling -------------- 0 12 15 24 # cumulative sum (i.e. delta) to be applied to each visit delta_df <- delta_template( impute_obj_CIR, delta = c(2, 2, 2, 2, 2, 2), dlag = c(1, 1, 1, 1, 1, 1) ) head(delta_df) #> id visit group is_mar is_missing is_post_ice strategy delta #> 1 id_1 1 Control TRUE TRUE TRUE MAR 2 #> 2 id_1 2 Control TRUE TRUE TRUE MAR 4 #> 3 id_1 3 Control TRUE TRUE TRUE MAR 6 #> 4 id_1 4 Control TRUE TRUE TRUE MAR 8 #> 5 id_1 5 Control TRUE TRUE TRUE MAR 10 #> 6 id_1 6 Control TRUE TRUE TRUE MAR 12 delta_df2 <- delta_df %>% mutate(delta = if_else(group == \"Control\", 0, delta)) head(delta_df2) #> id visit group is_mar is_missing is_post_ice strategy delta #> 1 id_1 1 Control TRUE TRUE TRUE MAR 0 #> 2 id_1 2 Control TRUE TRUE TRUE MAR 0 #> 3 id_1 3 Control TRUE TRUE TRUE MAR 0 #> 4 id_1 4 Control TRUE TRUE TRUE MAR 0 #> 5 id_1 5 Control TRUE TRUE TRUE MAR 0 #> 6 id_1 6 Control TRUE TRUE TRUE MAR 0 ana_delta <- analyse(impute_obj_CIR, delta = delta_df2, vars = vars_an) pool(ana_delta) #> #> Pool Object #> ----------- #> Number of Results Combined: 20 #> Method: rubin #> Confidence Level: 0.95 #> Alternative: two.sided #> #> Results: #> #> ================================================== #> parameter est se lci uci pval #> -------------------------------------------------- #> trt_1 -0.446 0.514 -1.459 0.567 0.386 #> lsm_ref_1 2.62 0.363 1.904 3.335 <0.001 #> lsm_alt_1 2.173 0.363 1.458 2.889 <0.001 #> trt_2 0.072 0.546 -1.006 1.15 0.895 #> lsm_ref_2 3.708 0.387 2.945 4.471 <0.001 #> lsm_alt_2 3.78 0.386 3.018 4.542 <0.001 #> trt_3 -1.507 0.626 -2.743 -0.272 0.017 #> lsm_ref_3 5.844 0.441 4.973 6.714 <0.001 #> lsm_alt_3 4.336 0.442 3.464 5.209 <0.001 #> trt_4 -2.062 0.731 -3.504 -0.621 0.005 #> lsm_ref_4 7.658 0.519 6.634 8.682 <0.001 #> lsm_alt_4 5.596 0.515 4.58 6.612 <0.001 #> trt_5 -2.938 0.916 -4.746 -1.13 0.002 #> lsm_ref_5 9.558 0.641 8.293 10.823 <0.001 #> lsm_alt_5 6.62 0.651 5.335 7.905 <0.001 #> trt_6 -3.53 1.045 -5.591 -1.469 0.001 #> lsm_ref_6 11.045 0.73 9.604 12.486 <0.001 #> lsm_alt_6 7.515 0.738 6.058 8.971 <0.001 #> --------------------------------------------------"},{"path":[]},{"path":"/articles/quickstart.html","id":"introduction","dir":"Articles","previous_headings":"","what":"Introduction","title":"rbmi: Quickstart","text":"purpose vignette provide 15 minute quickstart guide core functions rbmi package. rbmi package consists 4 core functions (plus several helper functions) typically called sequence: draws() - fits imputation models stores parameters impute() - creates multiple imputed datasets analyse() - analyses multiple imputed datasets pool() - combines analysis results across imputed datasets single statistic","code":""},{"path":"/articles/quickstart.html","id":"the-data","dir":"Articles","previous_headings":"","what":"The Data","title":"rbmi: Quickstart","text":"use publicly available example dataset antidepressant clinical trial active drug versus placebo. relevant endpoint Hamilton 17-item depression rating scale (HAMD17) assessed baseline weeks 1, 2, 4, 6. Study drug discontinuation occurred 24% subjects active drug 26% subjects placebo. data study drug discontinuation missing single additional intermittent missing observation. consider imputation model mean change baseline HAMD17 score outcome (variable CHANGE dataset). following covariates included imputation model: treatment group (THERAPY), (categorical) visit (VISIT), treatment--visit interactions, baseline HAMD17 score (BASVAL), baseline HAMD17 score--visit interactions. common unstructured covariance matrix structure assumed groups. analysis model ANCOVA model treatment group primary factor adjustment baseline HAMD17 score. rbmi expects input dataset complete; , must one row per subject visit. Missing outcome values coded NA, missing covariate values allowed. dataset incomplete, expand_locf() helper function can used add missing rows, using LOCF imputation carry forward observed baseline covariate values visits missing outcomes. Rows corresponding missing outcomes present antidepressant trial dataset. address therefore use expand_locf() function follows:","code":"library(rbmi) library(dplyr) #> #> Attaching package: 'dplyr' #> The following objects are masked from 'package:stats': #> #> filter, lag #> The following objects are masked from 'package:base': #> #> intersect, setdiff, setequal, union data(\"antidepressant_data\") dat <- antidepressant_data # Use expand_locf to add rows corresponding to visits with missing outcomes to the dataset dat <- expand_locf( dat, PATIENT = levels(dat$PATIENT), # expand by PATIENT and VISIT VISIT = levels(dat$VISIT), vars = c(\"BASVAL\", \"THERAPY\"), # fill with LOCF BASVAL and THERAPY group = c(\"PATIENT\"), order = c(\"PATIENT\", \"VISIT\") )"},{"path":"/articles/quickstart.html","id":"draws","dir":"Articles","previous_headings":"","what":"Draws","title":"rbmi: Quickstart","text":"draws() function fits imputation models stores corresponding parameter estimates Bayesian posterior parameter draws. three main inputs draws() function : data - primary longitudinal data.frame containing outcome variable covariates. data_ice - data.frame specifies first visit affected intercurrent event (ICE) imputation strategy handling missing outcome data ICE. one ICE imputed non-MAR strategy allowed per subject. method - statistical method used fit imputation models create imputed datasets. antidepressant trial data, dataset data_ice provided. However, can derived , dataset, subject’s first visit affected ICE “study drug discontinuation” corresponds first terminal missing observation. first derive dateset data_ice create 150 Bayesian posterior draws imputation model parameters. example, assume imputation strategy ICE Jump Reference (JR) subjects 150 multiple imputed datasets using Bayesian posterior draws imputation model created. Note use set_vars() specifies names key variables within dataset imputation model. Additionally, note whilst vars$group vars$visit added terms imputation model default, interaction , thus inclusion group * visit list covariates. Available imputation methods include: Bayesian multiple imputation - method_bayes() Approximate Bayesian multiple imputation - method_approxbayes() Conditional mean imputation (bootstrap) - method_condmean(type = \"bootstrap\") Conditional mean imputation (jackknife) - method_condmean(type = \"jackknife\") Bootstrapped multiple imputation - method = method_bmlmi() comparison methods, refer stat_specs vignette (Section 3.10). “statistical specifications” vignette (Section 3.10): vignette(\"stat_specs\",package=\"rbmi\"). Available imputation strategies include: Missing Random - \"MAR\" Jump Reference - \"JR\" Copy Reference - \"CR\" Copy Increments Reference - \"CIR\" Last Mean Carried Forward - \"LMCF\"","code":"# create data_ice and set the imputation strategy to JR for # each patient with at least one missing observation dat_ice <- dat %>% arrange(PATIENT, VISIT) %>% filter(is.na(CHANGE)) %>% group_by(PATIENT) %>% slice(1) %>% ungroup() %>% select(PATIENT, VISIT) %>% mutate(strategy = \"JR\") # In this dataset, subject 3618 has an intermittent missing values which does not correspond # to a study drug discontinuation. We therefore remove this subject from `dat_ice`. # (In the later imputation step, it will automatically be imputed under the default MAR assumption.) dat_ice <- dat_ice[-which(dat_ice$PATIENT == 3618),] dat_ice #> # A tibble: 43 × 3 #> PATIENT VISIT strategy #> #> 1 1513 5 JR #> 2 1514 5 JR #> 3 1517 5 JR #> 4 1804 7 JR #> 5 2104 7 JR #> 6 2118 5 JR #> 7 2218 6 JR #> 8 2230 6 JR #> 9 2721 5 JR #> 10 2729 5 JR #> # ℹ 33 more rows # Define the names of key variables in our dataset and # the covariates included in the imputation model using `set_vars()` # Note that the covariates argument can also include interaction terms vars <- set_vars( outcome = \"CHANGE\", visit = \"VISIT\", subjid = \"PATIENT\", group = \"THERAPY\", covariates = c(\"BASVAL*VISIT\", \"THERAPY*VISIT\") ) # Define which imputation method to use (here: Bayesian multiple imputation with 150 imputed datsets) method <- method_bayes( burn_in = 200, burn_between = 5, n_samples = 150, seed = 675442751 ) # Create samples for the imputation parameters by running the draws() function set.seed(987) drawObj <- draws( data = dat, data_ice = dat_ice, vars = vars, method = method, quiet = TRUE ) drawObj #> #> Draws Object #> ------------ #> Number of Samples: 150 #> Number of Failed Samples: 0 #> Model Formula: CHANGE ~ 1 + THERAPY + VISIT + BASVAL * VISIT + THERAPY * VISIT #> Imputation Type: random #> Method: #> name: Bayes #> burn_in: 200 #> burn_between: 5 #> same_cov: TRUE #> n_samples: 150 #> seed: 675442751"},{"path":"/articles/quickstart.html","id":"impute","dir":"Articles","previous_headings":"","what":"Impute","title":"rbmi: Quickstart","text":"next step use parameters imputation model generate imputed datasets. done via impute() function. function two key inputs: imputation model output draws() reference groups relevant reference-based imputation methods. ’s usage thus: instance, specifying PLACEBO group reference group well DRUG group (standard imputation using reference-based methods). Generally speaking, need see directly interact imputed datasets. However, wish inspect , can extracted imputation object using extract_imputed_dfs() helper function, .e.: Note case method_bayes() method_approxbayes(), imputed datasets correspond random imputations original dataset. method_condmean(), first imputed dataset always correspond completed original dataset containing subjects. method_condmean(type=\"jackknife\"), remaining datasets correspond conditional mean imputations leave-one-subject-datasets, whereas method_condmean(type=\"bootstrap\"), subsequent dataset corresponds conditional mean imputation bootstrapped datasets. method_bmlmi(), imputed datasets correspond sets random imputations bootstrapped datasets.","code":"imputeObj <- impute( drawObj, references = c(\"DRUG\" = \"PLACEBO\", \"PLACEBO\" = \"PLACEBO\") ) imputeObj #> #> Imputation Object #> ----------------- #> Number of Imputed Datasets: 150 #> Fraction of Missing Data (Original Dataset): #> 4: 0% #> 5: 8% #> 6: 13% #> 7: 25% #> References: #> DRUG -> PLACEBO #> PLACEBO -> PLACEBO imputed_dfs <- extract_imputed_dfs(imputeObj) head(imputed_dfs[[10]], 12) # first 12 rows of 10th imputed dataset #> PATIENT HAMATOTL PGIIMP RELDAYS VISIT THERAPY GENDER POOLINV BASVAL #> 1 new_pt_1 21 2 7 4 DRUG F 006 32 #> 2 new_pt_1 19 2 14 5 DRUG F 006 32 #> 3 new_pt_1 21 3 28 6 DRUG F 006 32 #> 4 new_pt_1 17 4 42 7 DRUG F 006 32 #> 5 new_pt_2 18 3 7 4 PLACEBO F 006 14 #> 6 new_pt_2 18 2 15 5 PLACEBO F 006 14 #> 7 new_pt_2 14 3 29 6 PLACEBO F 006 14 #> 8 new_pt_2 8 2 42 7 PLACEBO F 006 14 #> 9 new_pt_3 18 3 7 4 DRUG F 006 21 #> 10 new_pt_3 17 3 14 5 DRUG F 006 21 #> 11 new_pt_3 12 3 28 6 DRUG F 006 21 #> 12 new_pt_3 9 3 44 7 DRUG F 006 21 #> HAMDTL17 CHANGE #> 1 21 -11 #> 2 20 -12 #> 3 19 -13 #> 4 17 -15 #> 5 11 -3 #> 6 14 0 #> 7 9 -5 #> 8 5 -9 #> 9 20 -1 #> 10 18 -3 #> 11 16 -5 #> 12 13 -8"},{"path":"/articles/quickstart.html","id":"analyse","dir":"Articles","previous_headings":"","what":"Analyse","title":"rbmi: Quickstart","text":"next step run analysis model imputed dataset. done defining analysis function calling analyse() apply function imputed dataset. vignette use ancova() function provided rbmi package fits separate ANCOVA model outcomes visit returns treatment effect estimate corresponding least square means group per visit. Note , similar draws(), ancova() function uses set_vars() function determines names key variables within data covariates (addition treatment group) analysis model adjusted. Please also note names analysis estimates contain “ref” “alt” refer two treatment arms. particular “ref” refers first factor level vars$group necessarily coincide control arm. example, since levels(dat[[vars$group]]) = c(\"DRUG\", PLACEBO), results associated “ref” correspond intervention arm, associated “alt” correspond control arm. Additionally, can use delta argument analyse() perform delta adjustments imputed datasets prior analysis. brief, implemented specifying data.frame contains amount adjustment added longitudinal outcome subject visit, .e.  data.frame must contain columns subjid, visit, delta. appreciated carrying procedure potentially tedious, therefore delta_template() helper function provided simplify . particular, delta_template() returns shell data.frame delta-adjustment set 0 patients. Additionally delta_template() adds several meta-variables onto shell data.frame can used manual derivation manipulation delta-adjustment. example lets say want add delta-value 5 imputed values (.e. values missing original dataset) drug arm. implemented follows:","code":"anaObj <- analyse( imputeObj, ancova, vars = set_vars( subjid = \"PATIENT\", outcome = \"CHANGE\", visit = \"VISIT\", group = \"THERAPY\", covariates = c(\"BASVAL\") ) ) anaObj #> #> Analysis Object #> --------------- #> Number of Results: 150 #> Analysis Function: ancova #> Delta Applied: FALSE #> Analysis Estimates: #> trt_4 #> lsm_ref_4 #> lsm_alt_4 #> trt_5 #> lsm_ref_5 #> lsm_alt_5 #> trt_6 #> lsm_ref_6 #> lsm_alt_6 #> trt_7 #> lsm_ref_7 #> lsm_alt_7 # For reference show the additional meta variables provided delta_template(imputeObj) %>% as_tibble() #> # A tibble: 688 × 8 #> PATIENT VISIT THERAPY is_mar is_missing is_post_ice strategy delta #> #> 1 1503 4 DRUG TRUE FALSE FALSE NA 0 #> 2 1503 5 DRUG TRUE FALSE FALSE NA 0 #> 3 1503 6 DRUG TRUE FALSE FALSE NA 0 #> 4 1503 7 DRUG TRUE FALSE FALSE NA 0 #> 5 1507 4 PLACEBO TRUE FALSE FALSE NA 0 #> 6 1507 5 PLACEBO TRUE FALSE FALSE NA 0 #> 7 1507 6 PLACEBO TRUE FALSE FALSE NA 0 #> 8 1507 7 PLACEBO TRUE FALSE FALSE NA 0 #> 9 1509 4 DRUG TRUE FALSE FALSE NA 0 #> 10 1509 5 DRUG TRUE FALSE FALSE NA 0 #> # ℹ 678 more rows delta_df <- delta_template(imputeObj) %>% as_tibble() %>% mutate(delta = if_else(THERAPY == \"DRUG\" & is_missing , 5, 0)) %>% select(PATIENT, VISIT, delta) delta_df #> # A tibble: 688 × 3 #> PATIENT VISIT delta #> #> 1 1503 4 0 #> 2 1503 5 0 #> 3 1503 6 0 #> 4 1503 7 0 #> 5 1507 4 0 #> 6 1507 5 0 #> 7 1507 6 0 #> 8 1507 7 0 #> 9 1509 4 0 #> 10 1509 5 0 #> # ℹ 678 more rows anaObj_delta <- analyse( imputeObj, ancova, delta = delta_df, vars = set_vars( subjid = \"PATIENT\", outcome = \"CHANGE\", visit = \"VISIT\", group = \"THERAPY\", covariates = c(\"BASVAL\") ) )"},{"path":"/articles/quickstart.html","id":"pool","dir":"Articles","previous_headings":"","what":"Pool","title":"rbmi: Quickstart","text":"Finally, pool() function can used summarise analysis results across multiple imputed datasets provide overall statistic standard error, confidence intervals p-value hypothesis test null hypothesis effect equal 0. Note pooling method automatically derived based method specified original call draws(): method_bayes() method_approxbayes() pooling inference based Rubin’s rules. method_condmean(type = \"bootstrap\") inference either based normal approximation using bootstrap standard error (pool(..., type = \"normal\")) bootstrap percentiles (pool(..., type = \"percentile\")). method_condmean(type = \"jackknife\") inference based normal approximation using jackknife estimate standard error. method = method_bmlmi() inference according methods described von Hippel Bartlett (see stat_specs vignette details) Since used Bayesian multiple imputation vignette, pool() function automatically use Rubin’s rules. table values shown print message poolObj can also extracted using .data.frame() function: outputs gives estimated difference 2.077 (95% CI -0.133 4.287) two groups last visit associated p-value 0.065.","code":"poolObj <- pool( anaObj, conf.level = 0.95, alternative = \"two.sided\" ) poolObj #> #> Pool Object #> ----------- #> Number of Results Combined: 150 #> Method: rubin #> Confidence Level: 0.95 #> Alternative: two.sided #> #> Results: #> #> ================================================== #> parameter est se lci uci pval #> -------------------------------------------------- #> trt_4 -0.092 0.683 -1.439 1.256 0.893 #> lsm_ref_4 -1.616 0.486 -2.576 -0.656 0.001 #> lsm_alt_4 -1.708 0.475 -2.645 -0.77 <0.001 #> trt_5 1.278 0.927 -0.553 3.108 0.17 #> lsm_ref_5 -4.109 0.661 -5.415 -2.804 <0.001 #> lsm_alt_5 -2.831 0.647 -4.109 -1.553 <0.001 #> trt_6 1.907 0.997 -0.062 3.876 0.058 #> lsm_ref_6 -6.084 0.711 -7.49 -4.679 <0.001 #> lsm_alt_6 -4.177 0.698 -5.556 -2.799 <0.001 #> trt_7 2.077 1.118 -0.133 4.287 0.065 #> lsm_ref_7 -6.951 0.81 -8.552 -5.351 <0.001 #> lsm_alt_7 -4.875 0.796 -6.449 -3.3 <0.001 #> -------------------------------------------------- as.data.frame(poolObj) #> parameter est se lci uci pval #> 1 trt_4 -0.09180645 0.6826279 -1.43949684 1.2558839 8.931772e-01 #> 2 lsm_ref_4 -1.61581996 0.4862316 -2.57577141 -0.6558685 1.093708e-03 #> 3 lsm_alt_4 -1.70762640 0.4749573 -2.64531931 -0.7699335 4.262148e-04 #> 4 trt_5 1.27784334 0.9267958 -0.55262591 3.1083126 1.699077e-01 #> 5 lsm_ref_5 -4.10912984 0.6609343 -5.41453523 -2.8037244 4.327320e-09 #> 6 lsm_alt_5 -2.83128650 0.6469374 -4.10909149 -1.5534815 2.189243e-05 #> 7 trt_6 1.90697590 0.9965203 -0.06173754 3.8756893 5.753387e-02 #> 8 lsm_ref_6 -6.08439116 0.7114036 -7.48988902 -4.6788933 1.195687e-14 #> 9 lsm_alt_6 -4.17741526 0.6975145 -5.55556852 -2.7992620 1.482787e-08 #> 10 trt_7 2.07666647 1.1181891 -0.13345442 4.2867874 6.532284e-02 #> 11 lsm_ref_7 -6.95143063 0.8095397 -8.55204130 -5.3508200 1.622026e-14 #> 12 lsm_alt_7 -4.87476416 0.7963105 -6.44945627 -3.3000720 9.230722e-09"},{"path":"/articles/quickstart.html","id":"code","dir":"Articles","previous_headings":"","what":"Code","title":"rbmi: Quickstart","text":"report code presented vignette.","code":"library(rbmi) library(dplyr) data(\"antidepressant_data\") dat <- antidepressant_data # Use expand_locf to add rows corresponding to visits with missing outcomes to the dataset dat <- expand_locf( dat, PATIENT = levels(dat$PATIENT), # expand by PATIENT and VISIT VISIT = levels(dat$VISIT), vars = c(\"BASVAL\", \"THERAPY\"), # fill with LOCF BASVAL and THERAPY group = c(\"PATIENT\"), order = c(\"PATIENT\", \"VISIT\") ) # Create data_ice and set the imputation strategy to JR for # each patient with at least one missing observation dat_ice <- dat %>% arrange(PATIENT, VISIT) %>% filter(is.na(CHANGE)) %>% group_by(PATIENT) %>% slice(1) %>% ungroup() %>% select(PATIENT, VISIT) %>% mutate(strategy = \"JR\") # In this dataset, subject 3618 has an intermittent missing values which does not correspond # to a study drug discontinuation. We therefore remove this subject from `dat_ice`. # (In the later imputation step, it will automatically be imputed under the default MAR assumption.) dat_ice <- dat_ice[-which(dat_ice$PATIENT == 3618),] # Define the names of key variables in our dataset using `set_vars()` # and the covariates included in the imputation model # Note that the covariates argument can also include interaction terms vars <- set_vars( outcome = \"CHANGE\", visit = \"VISIT\", subjid = \"PATIENT\", group = \"THERAPY\", covariates = c(\"BASVAL*VISIT\", \"THERAPY*VISIT\") ) # Define which imputation method to use (here: Bayesian multiple imputation with 150 imputed datsets) method <- method_bayes( burn_in = 200, burn_between = 5, n_samples = 150, seed = 675442751 ) # Create samples for the imputation parameters by running the draws() function set.seed(987) drawObj <- draws( data = dat, data_ice = dat_ice, vars = vars, method = method, quiet = TRUE ) # Impute the data imputeObj <- impute( drawObj, references = c(\"DRUG\" = \"PLACEBO\", \"PLACEBO\" = \"PLACEBO\") ) # Fit the analysis model on each imputed dataset anaObj <- analyse( imputeObj, ancova, vars = set_vars( subjid = \"PATIENT\", outcome = \"CHANGE\", visit = \"VISIT\", group = \"THERAPY\", covariates = c(\"BASVAL\") ) ) # Apply a delta adjustment # Add a delta-value of 5 to all imputed values (i.e. those values # which were missing in the original dataset) in the drug arm. delta_df <- delta_template(imputeObj) %>% as_tibble() %>% mutate(delta = if_else(THERAPY == \"DRUG\" & is_missing , 5, 0)) %>% select(PATIENT, VISIT, delta) # Repeat the analyses with the adjusted values anaObj_delta <- analyse( imputeObj, ancova, delta = delta_df, vars = set_vars( subjid = \"PATIENT\", outcome = \"CHANGE\", visit = \"VISIT\", group = \"THERAPY\", covariates = c(\"BASVAL\") ) ) # Pool the results poolObj <- pool( anaObj, conf.level = 0.95, alternative = \"two.sided\" )"},{"path":"/articles/stat_specs.html","id":"scope-of-this-document","dir":"Articles","previous_headings":"","what":"Scope of this document","title":"rbmi: Statistical Specifications","text":"document describes statistical methods implemented rbmi R package standard reference-based multiple imputation continuous longitudinal outcomes. package implements three classes multiple imputation (MI) approaches: Conventional MI methods based Bayesian (approximate Bayesian) posterior draws model parameters combined Rubin’s rules make inferences described Carpenter, Roger, Kenward (2013) Cro et al. (2020). Conditional mean imputation methods combined re-sampling techniques described Wolbers et al. (2022). Bootstrapped MI methods described von Hippel Bartlett (2021). document structured follows: first provide informal introduction estimands corresponding treatment effect estimation based MI (section 2). core document consists section 3 describes statistical methodology detail also contains comparison implemented approaches (section 3.10). link theory functions included package rbmi described section 4. conclude comparison package alternative software implementations reference-based imputation methods (section 5).","code":""},{"path":[]},{"path":"/articles/stat_specs.html","id":"estimands","dir":"Articles","previous_headings":"2 Introduction to estimands and estimation methods","what":"Estimands","title":"rbmi: Statistical Specifications","text":"ICH E9(R1) addendum estimands sensitivity analyses describes systematic approach ensure alignment among clinical trial objectives, trial execution/conduct, statistical analyses, interpretation results (ICH E9 working group (2019)). per addendum, estimand precise description treatment effect reflecting clinical question posed trial objective summarizes population-level outcomes patients different treatment conditions compared. One important attribute estimand list possible intercurrent events (ICEs), .e. events occurring treatment initiation affect either interpretation existence measurements associated clinical question interest, definition appropriate strategies deal ICEs. three relevant strategies purpose document hypothetical strategy, treatment policy strategy, composite strategy. hypothetical strategy, scenario envisaged ICE occur. scenario, endpoint values ICE directly observable treated using models missing data. treatment policy strategy, treatment effect presence ICEs targeted analyses based observed outcomes regardless whether subject ICE . composite strategy, ICE included component endpoint.","code":""},{"path":"/articles/stat_specs.html","id":"alignment-between-the-estimand-and-the-estimation-method","dir":"Articles","previous_headings":"2 Introduction to estimands and estimation methods","what":"Alignment between the estimand and the estimation method","title":"rbmi: Statistical Specifications","text":"ICH E9(R1) addendum distinguishes ICEs missing data (ICH E9 working group (2019)). Whereas ICEs treatment discontinuations reflect clinical practice, amount missing data can minimized conduct clinical trial. However, many connections missing data ICEs. example, often difficult retain subjects clinical trial treatment discontinuation subject’s dropout trial leads missing data. another example, outcome values ICEs addressed using hypothetical strateg directly observable hypothetical scenario. Consequently, observed outcome values ICEs typically discarded treated missing data. addendum proposes estimation methods address problem presented missing data selected align estimand. recent overview methods align estimator estimand Mallinckrodt et al. (2020). short introduction estimation methods studies longitudinal endpoints can also found Wolbers et al. (2022). One prominent statistical method purpose multiple imputation (MI), target rbmi package.","code":""},{"path":"/articles/stat_specs.html","id":"missing-data-prior-to-ices","dir":"Articles","previous_headings":"2 Introduction to estimands and estimation methods > 2.2 Alignment between the estimand and the estimation method","what":"Missing data prior to ICEs","title":"rbmi: Statistical Specifications","text":"Missing data may occur subjects without ICE prior occurrence ICE. missing outcomes associated ICE, often plausible impute missing--random (MAR) assumption using standard MMRM imputation model longitudinal outcomes. Informally, MAR occurs missing data can fully accounted baseline variables included model observed longitudinal outcomes, model correctly specified.","code":""},{"path":"/articles/stat_specs.html","id":"implementation-of-the-hypothetical-strategy","dir":"Articles","previous_headings":"2 Introduction to estimands and estimation methods > 2.2 Alignment between the estimand and the estimation method","what":"Implementation of the hypothetical strategy","title":"rbmi: Statistical Specifications","text":"MAR imputation model described often also good starting point imputing data ICE handled using hypothetical strategy (Mallinckrodt et al. (2020)). Informally, assumes unobserved values ICE similar observed data subjects ICE remained follow-. However, situations, may reasonable assume missingness “informative” indicates systematically better worse outcome observed subjects. situations, MNAR imputation \\(\\delta\\)-adjustment explored sensitivity analysis. \\(\\delta\\)-adjustments add fixed random quantity imputations order make imputed outcomes systematically worse better observed described Cro et al. (2020). rbmi fixed \\(\\delta\\)-adjustments implemented.","code":""},{"path":"/articles/stat_specs.html","id":"implementation-of-the-treatment-policy-strategy","dir":"Articles","previous_headings":"2 Introduction to estimands and estimation methods > 2.2 Alignment between the estimand and the estimation method","what":"Implementation of the treatment policy strategy","title":"rbmi: Statistical Specifications","text":"Ideally, data collection continues ICE handled treatment policy strategy missing data arises. Indeed, post-ICE data increasingly systematically collected RCTs. However, despite best efforts, missing data ICE study treatment discontinuation may still occur subject drops study discontinuation. difficult give definite recommendations regarding implementation treatment policy strategy presence missing data stage optimal method highly context dependent topic ongoing statistical research. ICEs thought negligible effect efficacy outcomes, standard MAR-based imputation may appropriate. contrast, ICE treatment discontinuation may expected substantial impact efficacy outcomes. settings, MAR assumption may still plausible conditioning subject’s time-varying treatment status (Guizzaro et al. (2021)). case, one option impute missing post-discontinuation data based subjects also discontinued treatment continued followed (Polverejan Dragalin (2020)). Another option may require somewhat less post-discontinuation data include subjects imputation procedure model post-discontinuation data using time-varying treatment status indicators (e.g. time-varying indicators treatment compliance, discontinuation, initiation rescue treatment) (Guizzaro et al. (2021)). approach, post-ICE outcomes included every step analysis, including fitting imputation model. assumes ICEs may impact post-ICE outcomes otherwise missingness non-informative. approach also assumes time-varying covariates contain missing values, deviations outcomes ICE correctly modeled time-varying covariates, sufficient post-ICE data available inform regression coefficients time-varying covariates. proposals relatively recent remain open questions regarding appropriate trade-model complexity (e.g. model account potentially differential effect post-ICE outcomes depending timing ICE?) variance resulting treatment effect estimate. generally, yet established much post-discontinuation data required implement methods robustly without risk substantial inflation variance. trial settings, subjects discontinue randomized treatment. settings, treatment discontinuation rates higher difficult retain subjects trial treatment discontinuation leading sparse data collection treatment discontinuation. settings, amount available data treatment discontinuation may insufficient inform imputation model explicitly models post-discontinuation data. Depending disease area anticipated mechanism action intervention, may plausible assume subjects intervention group behave similarly subjects control group ICE treatment discontinuation. case, reference-based imputation methods option (Mallinckrodt et al. (2020)). Reference-based imputation methods formalize idea impute missing data intervention group based data control reference group. general description review reference-based imputation methods, refer Carpenter, Roger, Kenward (2013), Cro et al. (2020), . White, Royes, Best (2020) Wolbers et al. (2022). technical description implemented statistical methodology reference-based imputation, refer section 3 (particular section 3.4).","code":""},{"path":"/articles/stat_specs.html","id":"implementation-of-the-composite-strategy","dir":"Articles","previous_headings":"2 Introduction to estimands and estimation methods > 2.2 Alignment between the estimand and the estimation method","what":"Implementation of the composite strategy","title":"rbmi: Statistical Specifications","text":"composite strategy typically applied binary time--event outcomes can also used continuous outcomes ascribing suitably unfavorable value patients experience ICEs composite strategy defined. One possibility implement use MI \\(\\delta\\)-adjustment post-ICE data described Darken et al. (2020).","code":""},{"path":[]},{"path":"/articles/stat_specs.html","id":"sec:methodsOverview","dir":"Articles","previous_headings":"3 Statistical methodology","what":"Overview of the imputation procedure","title":"rbmi: Statistical Specifications","text":"Analyses datasets missing data always rely missing data assumptions. methods described can used produce valid imputations MAR assumption reference-based imputation assumptions. MNAR imputation based fixed \\(\\delta\\)-adjustments typically used sensitivity analyses tipping-point analyses also supported. Three general imputation approaches implemented rbmi: Conventional MI based Bayesian (approximate Bayesian) posterior draws imputation model combined Rubin’s rules inference described Carpenter, Roger, Kenward (2013) Cro et al. (2020). Conditional mean imputation based REML estimate imputation model combined resampling techniques (jackknife bootstrap) inference described Wolbers et al. (2022). Bootstrapped MI methods based REML estimates imputation model described von Hippel Bartlett (2021).","code":""},{"path":"/articles/stat_specs.html","id":"conventional-mi","dir":"Articles","previous_headings":"3 Statistical methodology > 3.1 Overview of the imputation procedure","what":"Conventional MI","title":"rbmi: Statistical Specifications","text":"Conventional MI approaches include following steps: Base imputation model fitting step (Section 3.3) Fit Bayesian multivariate normal mixed model repeated measures (MMRM) observed longitudinal outcomes exclusion data ICEs reference-based missing data imputation desired (Section 3.3.3). Draw \\(M\\) posterior samples estimated parameters (regression coefficients covariance matrices) model. Alternatively, \\(M\\) approximate posterior draws posterior distribution can sampled repeatedly applying conventional restricted maximum-likelihood (REML) parameter estimation MMRM model nonparametric bootstrap samples original dataset (Section 3.3.4). Imputation step (Section 3.4) Take single sample \\(m\\) (\\(m\\1,\\ldots, M)\\) posterior distribution imputation model parameters. subject, use sampled parameters defined imputation strategy determine mean covariance matrix describing subject’s marginal outcome distribution longitudinal outcome assessments (.e. observed missing outcomes). subjects, construct conditional multivariate normal distribution missing outcomes given observed outcomes (including observed outcomes ICEs reference-based assumption desired). subject, draw single sample conditional distribution impute missing outcomes leading complete imputed dataset. sensitivity analyses, pre-defined \\(\\delta\\)-adjustment may applied imputed data prior analysis step. (Section 3.5). Analysis step (Section 3.6) Analyze imputed dataset using analysis model (e.g. ANCOVA) resulting point estimate standard error (corresponding degrees freedom) treatment effect. Pooling step inference (Section 3.7) Repeat steps 2. 3. posterior sample \\(m\\), resulting \\(M\\) complete datasets, \\(M\\) point estimates treatment effect, \\(M\\) standard errors (corresponding degrees freedom). Pool \\(M\\) treatment effect estimates, standard errors, degrees freedom using rules Barnard Rubin obtain final pooled treatment effect estimator, standard error, degrees freedom.","code":""},{"path":"/articles/stat_specs.html","id":"conditional-mean-imputation","dir":"Articles","previous_headings":"3 Statistical methodology > 3.1 Overview of the imputation procedure","what":"Conditional mean imputation","title":"rbmi: Statistical Specifications","text":"conditional mean imputation approach includes following steps: Base imputation model fitting step (Section 3.3) Fit conventional multivariate normal/MMRM model using restricted maximum likelihood (REML) observed longitudinal outcomes exclusion data ICEs reference-based missing data imputation desired (Section 3.3.2). Imputation step (Section 3.4) subject, use fitted parameters step 1. construct conditional distribution missing outcomes given observed outcomes (including observed outcomes ICEs reference-based missing data imputation desired) described . subject, impute missing data deterministically mean conditional distribution leading complete imputed dataset. sensitivity analyses, pre-defined \\(\\delta\\)-adjustment may applied imputed data prior analysis step. (Section 3.5). Analysis step (Section 3.6) Apply analysis model (e.g. ANCOVA) completed dataset resulting point estimate treatment effect. Jackknife bootstrap inference step (Section 3.8) Inference treatment effect estimate 3. based re-sampling techniques. jackknife bootstrap supported. Importantly, methods require repeating steps imputation procedure (.e. imputation, conditional mean imputation, analysis steps) resampled datasets.","code":""},{"path":"/articles/stat_specs.html","id":"bootstrapped-mi","dir":"Articles","previous_headings":"3 Statistical methodology > 3.1 Overview of the imputation procedure","what":"Bootstrapped MI","title":"rbmi: Statistical Specifications","text":"bootstrapped MI approach includes following steps: Base imputation model fitting step (Section 3.3) Apply conventional restricted maximum-likelihood (REML) parameter estimation MMRM model \\(B\\) nonparametric bootstrap samples original dataset using observed longitudinal outcomes exclusion data ICEs reference-based missing data imputation desired. Imputation step (Section 3.4) Take bootstrapped dataset \\(b\\) (\\(b\\1,\\ldots, B)\\) corresponding imputation model parameter estimates. subject (bootstrapped dataset), use parameter estimates defined strategy dealing ICEs determine mean covariance matrix describing subject’s marginal outcome distribution longitudinal outcome assessments (.e. observed missing outcomes). subjects (bootstrapped dataset), construct conditional multivariate normal distribution missing outcomes given observed outcomes (including observed outcomes ICEs reference-based missing data imputation desired). subject (bootstrapped dataset), draw \\(D\\) samples conditional distributions impute missing outcomes leading \\(D\\) complete imputed dataset bootstrap sample \\(b\\). sensitivity analyses, pre-defined \\(\\delta\\)-adjustment may applied imputed data prior analysis step. (Section 3.5). Analysis step (Section 3.6) Analyze \\(B\\times D\\) imputed datasets using analysis model (e.g. ANCOVA) resulting \\(B\\times D\\) point estimates treatment effect. Pooling step inference (Section 3.9) Pool \\(B\\times D\\) treatment effect estimates described von Hippel Bartlett (2021) obtain final pooled treatment effect estimate, standard error, degrees freedom.","code":""},{"path":"/articles/stat_specs.html","id":"setting-notation-and-missing-data-assumptions","dir":"Articles","previous_headings":"3 Statistical methodology","what":"Setting, notation, and missing data assumptions","title":"rbmi: Statistical Specifications","text":"Assume data study \\(n\\) subjects total subject \\(\\) (\\(=1,\\ldots,n\\)) \\(J\\) scheduled follow-visits outcome interest assessed. applications, data randomized trial intervention vs control group treatment effect interest comparison outcomes specific visit randomized groups. However, single-arm trials multi-arm trials principle also supported rbmi implementation. Denote observed outcome vector length \\(J\\) subject \\(\\) \\(Y_i\\) (missing assessments coded NA (available)) non-missing missing components \\(Y_{!}\\) \\(Y_{?}\\), respectively. default, imputation missing outcomes \\(Y_{}\\) performed MAR assumption rbmi. Therefore, missing data following ICE handled using MAR imputation, compatible default assumption. discussed Section 2, MAR assumption often good starting point implementing hypothetical strategy. also note observed outcome data ICE handled using hypothetical strategy compatible strategy. Therefore, assume post-ICE data ICEs handled using hypothetical strategy already set NA \\(Y_i\\) prior calling rbmi functions. However, observed outcomes ICEs handled using treatment policy strategy included \\(Y_i\\) compatible strategy. Subjects may also experience one ICE missing data imputation according reference-based imputation method foreseen. subject \\(\\) ICE, denote first visit affected ICE \\(\\tilde{t}_i \\\\{1,\\ldots,J\\}\\). subjects, set \\(\\tilde{t}_i=\\infty\\). subject’s outcome vector setting observed outcomes visit \\(\\tilde{t}_i\\) onwards missing (.e. NA) denoted \\(Y'_i\\) corresponding data vector removal NA elements \\(Y'_{!}\\). MNAR \\(\\delta\\)-adjustments added imputed datasets formal imputation steps. covered separate section (Section 3.5).","code":""},{"path":[]},{"path":"/articles/stat_specs.html","id":"sec:imputationModelSpecs","dir":"Articles","previous_headings":"3 Statistical methodology > 3.3 The base imputation model","what":"Included data and model specification","title":"rbmi: Statistical Specifications","text":"purpose imputation model estimate (covariate-dependent) mean trajectories covariance matrices group absence ICEs handled using reference-based imputation methods. Conventionally, publications reference-based imputation methods implicitly assumed corresponding post-ICE data missing subjects (Carpenter, Roger, Kenward (2013)). also allow situation post-ICE data available subjects needs imputed using reference-based methods others. However, observed data ICEs reference-based imputation methods specified compatible imputation model described therefore removed considered missing purpose estimating imputation model, purpose . example, patient ICE addressed reference-based method outcomes ICE collected, post-ICE outcomes excluded fitting base imputation model (included following steps). , base imputation model fitted \\(Y'_{!}\\) \\(Y_{!}\\). exclude data, imputation model mistakenly estimate mean trajectories based mixture observed pre- post-ICE data relevant reference-based imputations. Observed post-ICE outcomes control reference group also excluded base imputation model user specifies reference-based imputation strategy ICEs. ensures ICE impact data included imputation model regardless whether ICE occurred control intervention group. hand, imputation reference group based MAR assumption even reference-based imputation methods may preferable settings include post-ICE data control group base imputation model. can implemented specifying MAR strategy ICE control group reference-based strategy ICE intervention group. base imputation model longitudinal outcomes \\(Y'_i\\) assumes mean structure linear function covariates. Full flexibility specification linear predictor model supported. minimum covariates include treatment group, (categorical) visit, treatment--visit interactions. Typically, covariates including baseline outcome also included. External time-varying covariates (e.g. calendar time visit) well internal time-varying (e.g. time-varying indicators treatment discontinuation initiation rescue treatment) may principle also included indicated (Guizzaro et al. (2021)). Missing covariate values allowed. means values time-varying covariates must non-missing every visit regardless whether outcome measured missing. Denote \\(J\\times p\\) design matrix subject \\(\\) corresponding mean structure model \\(X_i\\) matrix removal rows corresponding missing outcomes \\(Y'_{!}\\) \\(X'_{!}\\). \\(p\\) number parameters mean structure model elements \\(Y'_{!}\\). base imputation model observed outcomes defined : \\[ Y'_{!} = X'_{!}\\beta + \\epsilon_{!} \\mbox{ } \\epsilon_{!}\\sim N(0,\\Sigma_{!!})\\] \\(\\beta\\) vector regression coefficients \\(\\Sigma_{!!}\\) covariance matrix obtained complete-data \\(J\\times J\\)-covariance matrix \\(\\Sigma\\) omitting rows columns corresponding missing outcome assessments subject \\(\\). Typically, common unstructured covariance matrix subjects assumed \\(\\Sigma\\) separate covariate matrices per treatment group also supported. Indeed, implementation also supports specification separate covariate matrices according arbitrarily defined categorical variable groups subjects disjoint subset. example, useful different covariance matrices suspected different subject strata. Finally, imputation methods described rely Bayesian model fitting MCMC, flexibility choice covariance structure, .e. unstructured (default), heterogeneous Toeplitz, heterogeneous compound symmetry, AR(1) covariance structures supported.","code":""},{"path":"/articles/stat_specs.html","id":"sec:imputationModelREML","dir":"Articles","previous_headings":"3 Statistical methodology > 3.3 The base imputation model","what":"Restricted maximum likelihood estimation (REML)","title":"rbmi: Statistical Specifications","text":"Frequentist parameter estimation base imputation based REML. use REML improved alternative maximum likelihood (ML) covariance parameter estimation originally proposed Patterson Thompson (1971). Since , become default method parameter estimation linear mixed effects models. rbmi allows choose ML REML methods estimate model parameters, REML default option.","code":""},{"path":"/articles/stat_specs.html","id":"sec:imputationModelBayes","dir":"Articles","previous_headings":"3 Statistical methodology > 3.3 The base imputation model","what":"Bayesian model fitting","title":"rbmi: Statistical Specifications","text":"Bayesian imputation model fitted R package rstan (Stan Development Team (2020)). rstan R interface Stan. Stan powerful flexible statistical software developed dedicated team implements Bayesian inference state---art MCMC sampling procedures. multivariate normal model missing data specified section 3.3.1 can considered generalization models described Stan user’s guide (see Stan Development Team (2020, sec. 3.5)). prior distributions SAS implementation “five macros” used (Roger (2021)), .e. improper flat priors regression coefficients weakly informative inverse Wishart prior covariance matrix (matrices). Specifically, let \\(S \\\\mathbb{R}^{J \\times J}\\) symmetric positive definite matrix \\(\\nu \\(J-1, \\infty)\\). symmetric positive definite matrix \\(x \\\\mathbb{R}^{J \\times J}\\) density: \\[ \\text{InvWish}(x \\vert \\nu, S) = \\frac{1}{2^{\\nu J/2}} \\frac{1}{\\Gamma_J(\\frac{\\nu}{2})} \\vert S \\vert^{\\nu/2} \\vert x \\vert ^{-(\\nu + J + 1)/2} \\text{exp}(-\\frac{1}{2} \\text{tr}(Sx^{-1})). \\] \\(\\nu > J+1\\) mean given : \\[ E[x] = \\frac{S}{\\nu - J - 1}. \\] choose \\(S\\) equal estimated covariance matrix frequentist REML fit \\(\\nu = J+2\\) lowest degrees freedom guarantee finite mean. Setting degrees freedom low \\(\\nu\\) ensures prior little impact posterior. Moreover, choice allows interpret parameter \\(S\\) mean prior distribution. “five macros”, MCMC algorithm initialized parameters frequentist REML fit (see section 3.3.2). described , using weakly informative priors parameters. Therefore, Markov chain essentially starting targeted stationary posterior distribution minimal amount burn-chain required.","code":""},{"path":"/articles/stat_specs.html","id":"sec:imputationModelBoot","dir":"Articles","previous_headings":"3 Statistical methodology > 3.3 The base imputation model","what":"Approximate Bayesian posterior draws via the bootstrap","title":"rbmi: Statistical Specifications","text":"Several authors suggested stabler way get Bayesian posterior draws imputation model bootstrap incomplete data calculate REML estimates bootstrap sample (Little Rubin (2002), Efron (1994), Honaker King (2010), von Hippel Bartlett (2021)). method proper REML estimates bootstrap samples asymptotically equivalent sample posterior distribution may provide additional robustness model misspecification (Little Rubin (2002, sec. 10.2.3, part 6), Honaker King (2010)). order retain balance treatment groups stratification factors across bootstrap samples, user able provide stratification variables bootstrap rbmi implementation.","code":""},{"path":[]},{"path":"/articles/stat_specs.html","id":"sec:imputatioMNAR","dir":"Articles","previous_headings":"3 Statistical methodology > 3.4 Imputation step","what":"Marginal imputation distribution for a subject - MAR case","title":"rbmi: Statistical Specifications","text":"subject \\(\\), marginal distribution complete \\(J\\)-dimensional outcome vector assessment visits according imputation model multivariate normal distribution. mean \\(\\tilde{\\mu}_i\\) given predicted mean imputation model conditional subject’s baseline characteristics, group, , optionally, time-varying covariates. covariance matrix \\(\\tilde{\\Sigma}_i\\) given overall estimated covariance matrix , different covariance matrices assumed different groups, covariance matrix corresponding subject \\(\\)’s group.","code":""},{"path":"/articles/stat_specs.html","id":"sec:imputationRefBased","dir":"Articles","previous_headings":"3 Statistical methodology > 3.4 Imputation step","what":"Marginal imputation distribution for a subject - reference-based imputation methods","title":"rbmi: Statistical Specifications","text":"subject \\(\\), calculate mean covariance matrix complete \\(J\\)-dimensional outcome vector assessment visits MAR case denote \\(\\mu_i\\) \\(\\Sigma_i\\). reference-based imputation methods, corresponding reference group also required group. Typically, reference group intervention group control group. reference mean \\(\\mu_{ref,}\\) defined predicted mean imputation model conditional reference group (rather actual group subject \\(\\) belongs ) subject’s baseline characteristics. reference covariance matrix \\(\\Sigma_{ref,}\\) overall estimated covariance matrix , different covariance matrices assumed different groups, estimated covariance matrix corresponding reference group. principle, time-varying covariates also included reference-based imputation methods. However, sensible external time-varying covariates (e.g. calendar time visit) internal time-varying covariates (e.g. treatment discontinuation) latter likely depend actual treatment group typically sensible assume trajectory time-varying covariate reference group. Based means covariance matrices, subject’s marginal imputation distribution reference-based imputation methods calculated detailed Carpenter, Roger, Kenward (2013, sec. 4.3). Denote mean covariance matrix marginal imputation distribution \\(\\tilde{\\mu}_i\\) \\(\\tilde{\\Sigma}_i\\). Recall subject’s first visit affected ICE denoted \\(\\tilde{t}_i \\\\{1,\\ldots,J\\}\\) (visit \\(\\tilde{t}_i-1\\) last visit unaffected ICE). marginal distribution patient \\(\\) built according specific assumption data post ICE follows: Jump reference (JR): patient’s outcome distribution normally distributed following mean: \\[\\tilde{\\mu}_i = (\\mu_i[1], \\dots, \\mu_i[\\tilde{t}_i-1], \\mu_{ref,}[\\tilde{t}_i], \\dots, \\mu_{ref,}[J])^T.\\] covariance matrix constructed follows. First, partition covariance matrices \\(\\Sigma_i\\) \\(\\Sigma_{ref,}\\) blocks according time ICE \\(\\tilde{t}_i\\): \\[ \\Sigma_{} = \\begin{bmatrix} \\Sigma_{, 11} & \\Sigma_{, 12} \\\\ \\Sigma_{, 21} & \\Sigma_{,22} \\\\ \\end{bmatrix} \\] \\[ \\Sigma_{ref,} = \\begin{bmatrix} \\Sigma_{ref, , 11} & \\Sigma_{ref, , 12} \\\\ \\Sigma_{ref, , 21} & \\Sigma_{ref, ,22} \\\\ \\end{bmatrix}. \\] want covariance matrix \\(\\tilde{\\Sigma}_i\\) match \\(\\Sigma_i\\) pre-deviation measurements, \\(\\Sigma_{ref,}\\) conditional components post-deviation given pre-deviation measurements. solution derived Carpenter, Roger, Kenward (2013, sec. 4.3) given : \\[ \\begin{matrix} \\tilde{\\Sigma}_{,11} = \\Sigma_{, 11} \\\\ \\tilde{\\Sigma}_{, 21} = \\Sigma_{ref,, 21} \\Sigma^{-1}_{ref,, 11} \\Sigma_{, 11} \\\\ \\tilde{\\Sigma}_{, 22} = \\Sigma_{ref, , 22} - \\Sigma_{ref,, 21} \\Sigma^{-1}_{ref,, 11} (\\Sigma_{ref,, 11} - \\Sigma_{,11}) \\Sigma^{-1}_{ref,, 11} \\Sigma_{ref,, 12}. \\end{matrix} \\] Copy increments reference (CIR): patient’s outcome distribution normally distributed following mean: \\[ \\begin{split} \\tilde{\\mu}_i =& (\\mu_i[1], \\dots, \\mu_i[\\tilde{t}_i-1], \\mu_i[\\tilde{t}_i-1] + (\\mu_{ref,}[\\tilde{t}_i] - \\mu_{ref,}[\\tilde{t}_i-1]), \\dots,\\\\ & \\mu_i[\\tilde{t}_i-1]+(\\mu_{ref,}[J] - \\mu_{ref,}[\\tilde{t}_i-1]))^T. \\end{split} \\] covariance matrix derived JR method. Copy reference (CR): patient’s outcome distribution normally distributed mean covariance matrix taken reference group: \\[ \\tilde{\\mu}_i = \\mu_{ref,} \\] \\[ \\tilde{\\Sigma}_i = \\Sigma_{ref,}. \\] Last mean carried forward (LMCF): patient’s outcome distribution normally distributed following mean: \\[ \\tilde{\\mu}_i = (\\mu_i[1], \\dots, \\mu_i[\\tilde{t}_i-1], \\mu_i[\\tilde{t}_i-1], \\dots, \\mu_i[\\tilde{t}_i-1])'\\] covariance matrix: \\[ \\tilde{\\Sigma}_i = \\Sigma_i.\\]","code":""},{"path":"/articles/stat_specs.html","id":"sec:imputationRandomConditionalMean","dir":"Articles","previous_headings":"3 Statistical methodology > 3.4 Imputation step","what":"Imputation of missing outcome data","title":"rbmi: Statistical Specifications","text":"joint marginal multivariate normal imputation distribution subject \\(\\)’s observed missing outcome data mean \\(\\tilde{\\mu}_i\\) covariance matrix \\(\\tilde{\\Sigma}_i\\) defined . actual imputation missing outcome data obtained conditioning marginal distribution subject’s observed outcome data. note, approach valid regardless whether subject intermittent terminal missing data. conditional distribution used imputation multivariate normal distribution explicit formulas conditional mean covariance readily available. completeness, report notation terminology setting. marginal distribution outcome patient \\(\\) \\(Y_i \\sim N(\\tilde{\\mu}_i, \\tilde{\\Sigma}_i)\\) outcome \\(Y_i\\) can decomposed observed (\\(Y_{,!}\\)) unobserved (\\(Y_{,?}\\)) components. Analogously mean \\(\\tilde{\\mu}_i\\) can decomposed \\((\\tilde{\\mu}_{,!},\\tilde{\\mu}_{,?})\\) covariance \\(\\tilde{\\Sigma}_i\\) : \\[ \\tilde{\\Sigma}_i = \\begin{bmatrix} \\tilde{\\Sigma}_{, !!} & \\tilde{\\Sigma}_{,!?} \\\\ \\tilde{\\Sigma}_{, ?!} & \\tilde{\\Sigma}_{, ??} \\end{bmatrix}. \\] conditional distribution \\(Y_{,?}\\) conditional \\(Y_{,!}\\) multivariate normal distribution expectation \\[ E(Y_{,?} \\vert Y_{,!})= \\tilde{\\mu}_{,?} + \\tilde{\\Sigma}_{, ?!} \\tilde{\\Sigma}_{,!!}^{-1} (Y_{,!} - \\tilde{\\mu}_{,!}) \\] covariance matrix \\[ Cov(Y_{,?} \\vert Y_{,!}) = \\tilde{\\Sigma}_{,??} - \\tilde{\\Sigma}_{,?!} \\tilde{\\Sigma}_{,!!}^{-1} \\tilde{\\Sigma}_{,!?}. \\] Conventional random imputation consists sampling conditional multivariate normal distribution. Conditional mean imputation imputes missing values deterministic conditional expectation \\(E(Y_{,?} \\vert Y_{,!})\\).","code":""},{"path":"/articles/stat_specs.html","id":"sec:deltaAdjustment","dir":"Articles","previous_headings":"3 Statistical methodology","what":"\\(\\delta\\)-adjustment","title":"rbmi: Statistical Specifications","text":"marginal \\(\\delta\\)-adjustment approach similar “five macros” SAS implemented (Roger (2021)), .e. fixed non-stochastic values added multivariate normal imputation step prior analysis. relevant sensitivity analyses order make imputed data systematically worse better, respectively, observed data. addition, authors suggested \\(\\delta\\)-type adjustments implement composite strategy continuous outcomes (Darken et al. (2020)). implementation provides full flexibility regarding specific implementation \\(\\delta\\)-adjustment, .e. value added may depend randomized treatment group, timing subject’s ICE, factors. suggestions case studies regarding topic, refer Cro et al. (2020).","code":""},{"path":"/articles/stat_specs.html","id":"sec:analysis","dir":"Articles","previous_headings":"3 Statistical methodology","what":"Analysis step","title":"rbmi: Statistical Specifications","text":"data imputation, standard analysis model can applied completed data resulting treatment effect estimate. imputed data longer contains missing values, analysis model often simple. example, can analysis covariance (ANCOVA) model outcome (change outcome baseline) specific visit j dependent variable, randomized treatment group primary covariate , typically, adjustment baseline covariates imputation model.","code":""},{"path":"/articles/stat_specs.html","id":"sec:pooling","dir":"Articles","previous_headings":"3 Statistical methodology","what":"Pooling step for inference of (approximate) Bayesian MI and Rubin’s rules","title":"rbmi: Statistical Specifications","text":"Assume analysis model applied \\(M\\) multiple imputed random datasets resulted \\(m\\) treatment effect estimates \\(\\hat{\\theta}_m\\) (\\(m=1,\\ldots,M\\)) corresponding standard error \\(SE_m\\) (available) degrees freedom \\(\\nu_{com}\\). degrees freedom available analysis model, set \\(\\nu_{com}=\\infty\\) inference based normal distribution. Rubin’s rules used pooling treatment effect estimates corresponding variances estimates analysis steps across \\(M\\) multiple imputed datasets. According Rubin’s rules, final estimate treatment effect calculated sample mean \\(M\\) treatment effect estimates: \\[ \\hat{\\theta} = \\frac{1}{M} \\sum_{m = 1}^M \\hat{\\theta}_m. \\] pooled variance based two components reflect within variance treatment effects across multiple imputed datasets: \\[ V(\\hat{\\theta}) = V_W(\\hat{\\theta}) + (1 + \\frac{1}{M}) V_B(\\hat{\\theta}) \\] \\(V_W(\\hat{\\theta}) = \\frac{1}{M}\\sum_{m = 1}^M SE^2_m\\) within-variance \\(V_B(\\hat{\\theta}) = \\frac{1}{M-1} \\sum_{m = 1}^M (\\hat{\\theta}_m - \\hat{\\theta})^2\\) -variance. Confidence intervals tests null hypothesis \\(H_0: \\theta=\\theta_0\\) based \\(t\\)-statistics \\(T\\): \\[ T= (\\hat{\\theta}-\\theta_0)/\\sqrt{V(\\hat{\\theta})}. \\] null hypothesis, \\(T\\) approximate \\(t\\)-distribution \\(\\nu\\) degrees freedom. \\(\\nu\\) calculated according Barnard Rubin approximation, see Barnard Rubin (1999) (formula 3) Little Rubin (2002) (formula (5.24), page 87): \\[ \\nu = \\frac{\\nu_{old}* \\nu_{obs}}{\\nu_{old} + \\nu_{obs}} \\] \\[ \\nu_{old} = \\frac{M-1}{\\lambda^2} \\quad\\mbox{}\\quad \\nu_{obs} = \\frac{\\nu_{com} + 1}{\\nu_{com} + 3} \\nu_{com} (1 - \\lambda) \\] \\(\\lambda = \\frac{(1 + \\frac{1}{M})V_B(\\hat{\\theta})}{V(\\hat{\\theta})}\\) fraction missing information.","code":""},{"path":[]},{"path":"/articles/stat_specs.html","id":"point-estimate-of-the-treatment-effect","dir":"Articles","previous_headings":"3 Statistical methodology > 3.8 Bootstrap and jackknife inference for conditional mean imputation","what":"Point estimate of the treatment effect","title":"rbmi: Statistical Specifications","text":"point estimator obtained applying analysis model (Section 3.6) single conditional mean imputation missing data (see Section 3.4.3) based REML estimator parameters imputation model (see Section 3.3.2). denote treatment effect estimator \\(\\hat{\\theta}\\). demonstrated Wolbers et al. (2022) (Section 2.4), treatment effect estimator valid analysis model ANCOVA model , generally, treatment effect estimator linear function imputed outcome vector. Indeed, case, estimator identical pooled treatment effect across multiple random REML imputation infinite number imputations corresponds computationally efficient implementation proposal von Hippel Bartlett (2021). expect conditional mean imputation method also applicable analysis models (e.g. general MMRM analysis models) formally justified.","code":""},{"path":"/articles/stat_specs.html","id":"jackknife-standard-errors-confidence-intervals-ci-and-tests-for-the-treatment-effect","dir":"Articles","previous_headings":"3 Statistical methodology > 3.8 Bootstrap and jackknife inference for conditional mean imputation","what":"Jackknife standard errors, confidence intervals (CI) and tests for the treatment effect","title":"rbmi: Statistical Specifications","text":"dataset containing \\(n\\) subjects, jackknife standard error depends treatment effect estimates \\(\\hat{\\theta}_{(-b)}\\) (\\(b=1,\\ldots,n\\)) samples original dataset leave observation subject \\(b\\). described previously, obtain treatment effect estimates leave-one-subject-datasets, steps imputation procedure (.e. imputation, conditional mean imputation, analysis steps) need repeated new dataset. , jackknife standard error defined \\[\\hat{se}_{jack}=[\\frac{(n-1)}{n}\\cdot\\sum_{b=1}^{n} (\\hat{\\theta}_{(-b)}-\\bar{\\theta}_{(.)})^2]^{1/2}\\] \\(\\bar{\\theta}_{(.)}\\) denotes mean jackknife estimates (Efron Tibshirani (1994), chapter 10). corresponding two-sided normal approximation \\(1-\\alpha\\) CI defined \\(\\hat{\\theta}\\pm z^{1-\\alpha/2}\\cdot \\hat{se}_{jack}\\) \\(\\hat{\\theta}\\) treatment effect estimate original dataset. Tests null hypothesis \\(H_0: \\theta=\\theta_0\\) based \\(Z\\)-score \\(Z=(\\hat{\\theta}-\\theta_0)/\\hat{se}_{jack}\\) using standard normal approximation. simulation study reported Wolbers et al. (2022) demonstrated exact protection type error jackknife-based inference relatively low sample size (n = 100 per group) substantial amount missing data (>25% subjects ICE).","code":""},{"path":"/articles/stat_specs.html","id":"bootstrap-standard-errors-confidence-intervals-ci-and-tests-for-the-treatment-effect","dir":"Articles","previous_headings":"3 Statistical methodology > 3.8 Bootstrap and jackknife inference for conditional mean imputation","what":"Bootstrap standard errors, confidence intervals (CI) and tests for the treatment effect","title":"rbmi: Statistical Specifications","text":"alternative jackknife, bootstrap also implemented rbmi (Efron Tibshirani (1994), Davison Hinkley (1997)). Two different bootstrap methods implemented rbmi: Methods based bootstrap standard error normal approximation percentile bootstrap methods. Denote treatment effect estimates \\(B\\) bootstrap samples \\(\\hat{\\theta}^*_b\\) (\\(b=1,\\ldots,B\\)). bootstrap standard error \\(\\hat{se}_{boot}\\) defined empirical standard deviation bootstrapped treatment effect estimates. Confidence intervals tests based bootstrap standard error can constructed way jackknife. Confidence intervals using percentile bootstrap based empirical quantiles bootstrap distribution corresponding statistical tests implemented rbmi via inversion confidence interval. Explicit formulas bootstrap inference implemented rbmi package considerations regarding required number bootstrap samples included Appendix Wolbers et al. (2022). simulation study reported Wolbers et al. (2022) demonstrated small inflation type error rate inference based bootstrap standard error (\\(5.3\\%\\) nominal type error rate \\(5\\%\\)) sample size n = 100 per group substantial amount missing data (>25% subjects ICE). Based simulations, recommend jackknife bootstrap inference performed better simulation study typically much faster compute bootstrap.","code":""},{"path":"/articles/stat_specs.html","id":"sec:poolbmlmi","dir":"Articles","previous_headings":"3 Statistical methodology","what":"Pooling step for inference of the bootstrapped MI methods","title":"rbmi: Statistical Specifications","text":"Assume analysis model applied \\(B\\times D\\) multiple imputed random datasets resulted \\(B\\times D\\) treatment effect estimates \\(\\hat{\\theta}_{bd}\\) (\\(b=1,\\ldots,B\\); \\(d=1,\\ldots,D\\)). final estimate treatment effect calculated sample mean \\(B*D\\) treatment effect estimates: \\[ \\hat{\\theta} = \\frac{1}{BD} \\sum_{b = 1}^B \\sum_{d = 1}^D \\hat{\\theta}_{bd}. \\] pooled variance based two components reflect variability within imputed bootstrap samples (von Hippel Bartlett (2021), formula 8.4): \\[ V(\\hat{\\theta}) = (1 + \\frac{1}{B})\\frac{MSB - MSW}{D} + \\frac{MSW}{BD} \\] \\(MSB\\) mean square bootstrapped datasets, \\(MSW\\) mean square within bootstrapped datasets imputed datasets: \\[ \\begin{align*} MSB &= \\frac{D}{B-1} \\sum_{b = 1}^B (\\bar{\\theta_{b}} - \\hat{\\theta})^2 \\\\ MSW &= \\frac{1}{B(D-1)} \\sum_{b = 1}^B \\sum_{d = 1}^D (\\theta_{bd} - \\bar{\\theta_b})^2 \\end{align*} \\] \\(\\bar{\\theta_{b}}\\) mean across \\(D\\) estimates obtained random imputation \\(b\\)-th bootstrap sample. degrees freedom estimated following formula (von Hippel Bartlett (2021), formula 8.6): \\[ \\nu = \\frac{(MSB\\cdot (B+1) - MSW\\cdot B)^2}{\\frac{MSB^2\\cdot (B+1)^2}{B-1} + \\frac{MSW^2\\cdot B}{D-1}} \\] Confidence intervals tests null hypothesis \\(H_0: \\theta=\\theta_0\\) based \\(t\\)-statistics \\(T\\): \\[ T= (\\hat{\\theta}-\\theta_0)/\\sqrt{V(\\hat{\\theta})}. \\] null hypothesis, \\(T\\) approximate \\(t\\)-distribution \\(\\nu\\) degrees freedom.","code":""},{"path":[]},{"path":"/articles/stat_specs.html","id":"treatment-effect-estimation","dir":"Articles","previous_headings":"3 Statistical methodology > 3.10 Comparison between the implemented approaches","what":"Treatment effect estimation","title":"rbmi: Statistical Specifications","text":"approaches provide consistent treatment effect estimates standard reference-based imputation methods case analysis model completed datasets general linear model ANCOVA. Methods conditional mean imputation also valid analysis models. validity conditional mean imputation formally demonstrated analyses using general linear model (Wolbers et al. (2022, sec. 2.4)) though may also applicable widely (e.g. general MMRM analysis models). Treatment effects based conditional mean imputation deterministic. methods affected Monte Carlo sampling error precision estimates depends number imputations bootstrap samples, respectively.","code":""},{"path":"/articles/stat_specs.html","id":"standard-errors-of-the-treatment-effect","dir":"Articles","previous_headings":"3 Statistical methodology > 3.10 Comparison between the implemented approaches","what":"Standard errors of the treatment effect","title":"rbmi: Statistical Specifications","text":"approaches imputation MAR assumption provide consistent estimates frequentist standard error. reference-based imputation methods, situation complicated two different types variance estimators proposed statistical literature (Bartlett (2023)). first frequentist variance describes actual repeated sampling variability estimator. reference-based missing data assumption correctly specified, resulting inference based variance correct frequentist sense, .e. hypothesis tests asymptotically correct type error control confidence intervals correct coverage probabilities repeated sampling (Bartlett (2023), Wolbers et al. (2022)). Reference-based missing data assumptions strong borrow information reference arm imputation active arm. consequence, size frequentist standard errors treatment effects may decrease increasing amounts missing data. second proposal -called “information-anchored” variance originally proposed context sensitivity analyses (Cro, Carpenter, Kenward (2019)). variance estimator based disentangling point estimation variance estimation altogether. information-anchoring principle described Cro, Carpenter, Kenward (2019) states relative increase variance treatment effect estimator MAR imputation increasing amounts missing data preserved reference-based imputation methods. resulting information-anchored variance typically similar variance MAR imputation typically increases increasing amounts missing data. However, information-anchored variance reflect actual variability reference-based estimator repeated sampling resulting inference highly conservative resulting substantial power loss (Wolbers et al. (2022)). Moreover, date, Bayesian frequentist framework developed information-anchored variance provides correct inference reference-based missingness assumptions, clear whether framework can even developed. Reference-based conditional mean imputation (method_condmean()) bootstrapped likelihood-based multiple methods (method = method_bmlmi()) obtain standard errors via resampling hence target frequentist variance (Wolbers et al. (2022), von Hippel Bartlett (2021)). finite samples, simulations sample size \\(n=100\\) per group reported Wolbers et al. (2022) demonstrated conditional mean imputation combined jackknife (method_condmean(type = \"jackknife\")) provided exact protection type one error rate whereas bootstrap (method_condmean(type = \"bootstrap\")) associated small type error inflation (5.1% 5.3% nominal level 5%). reference-based conditional mean imputation, alternative information-anchored variance can obtained following proposal Lu (2021). basic idea Lu (2021) obtain information-anchored variance via MAR imputation combined delta-adjustment delta selected data-driven way match reference-based estimator. conditional mean imputation, proposal Lu (2021) can implemented choosing delta-adjustment difference conditional mean imputation chosen reference-based assumption MAR original dataset. illustration different variances can obtained conditional mean imputation rbmi provided vignette “Frequentist information-anchored inference reference-based conditional mean imputation” (vignette(topic = \"CondMean_Inference\", package = \"rbmi\")). Reference-based Bayesian (approximate Bayesian) multiple imputation methods combined Rubin’s rules (method_bayes() method_approxbayes()) target information-anchored variance (Cro, Carpenter, Kenward (2019)). frequentist variance methods principle obtained via bootstrap jackknife re-sampling treatment effect estimates computationally intensive directly supported rbmi. view primary analyses, accurate type error control (can obtained using frequentist variance) important adherence information anchoring principle , us, fully compatible strong reference-based assumptions. case, reference-based imputation used primary analysis, critical chosen reference-based assumption can clinically justified, suitable sensitivity analyses conducted stress-test assumptions. Conditional mean imputation combined jackknife method leads deterministic standard error estimates , consequently, confidence intervals \\(p\\)-values also deterministic. particularly important regulatory setting important ascertain whether calculated \\(p\\)-value close critical boundary 5% truly threshold rather uncertain Monte Carlo error.","code":""},{"path":"/articles/stat_specs.html","id":"computational-complexity","dir":"Articles","previous_headings":"3 Statistical methodology > 3.10 Comparison between the implemented approaches","what":"Computational complexity","title":"rbmi: Statistical Specifications","text":"Bayesian MI methods rely specification prior distributions usage Markov chain Monte Carlo (MCMC) methods. methods based multiple imputation bootstrapping require tuning parameters specification number imputations \\(M\\) bootstrap samples \\(B\\) rely numerical optimization fitting MMRM imputation models via REML. Conditional mean imputation combined jackknife tuning parameters. rbmi implementation, fitting MMRM imputation model via REML computationally expensive. MCMC sampling using rstan (Stan Development Team (2020)) typically relatively fast setting requires small burn-burn-chains. addition, number random imputations reliable inference using Rubin’s rules often smaller number resamples required jackknife bootstrap (see e.g. discussions . R. White, Royston, Wood (2011, sec. 7) Bayesian MI Appendix Wolbers et al. (2022) bootstrap). Thus, many applications, expect conventional MI based Bayesian posterior draws fastest, followed conventional MI using approximate Bayesian posterior draws conditional mean imputation combined jackknife. Conditional mean imputation combined bootstrap bootstrapped MI methods typically computationally demanding. note, implemented methods conceptually straightforward parallelise parallelisation support provided rbmi.","code":""},{"path":"/articles/stat_specs.html","id":"sec:rbmiFunctions","dir":"Articles","previous_headings":"","what":"Mapping of statistical methods to rbmi functions","title":"rbmi: Statistical Specifications","text":"full documentation rbmi package functionality refer help pages functions package vignettes. give brief overview different steps imputation procedure mapped rbmi functions: Bayesian posterior parameter draws imputation model obtained via argument method = method_bayes(). Approximate Bayesian posterior parameter draws imputation model obtained via argument method = method_approxbayes(). ML REML parameter estimates imputation model parameters original dataset leave-one-subject-datasets (required jackknife) obtained via argument method = method_condmean(type = \"jackknife\"). ML REML parameter estimates imputation model parameters original dataset bootstrapped datasets obtained via argument method = method_condmean(type = \"bootstrap\"). Bootstrapped MI methods obtained via argument method = method_bmlmi(B=B, D=D) \\(B\\) refers number bootstrap samples \\(D\\) number random imputations bootstrap sample. imputation step using random imputation deterministic conditional mean imputation, respectively, implemented function impute(). Imputation can performed assuming already implemented imputation strategies presented section 3.4. Additionally, user-defined imputation strategies also supported. analysis step implemented function analyse() applies analysis model imputed datasets. default, analysis model (argument fun) ancova() function alternative analysis functions can also provided user. analyse() function also allows \\(\\delta\\)-adjustments imputed datasets prior analysis via argument delta. inference step implemented function pool() pools results across imputed datasets. Rubin Bernard rule applied case (approximate) Bayesian MI. conditional mean imputation, jackknife bootstrap (normal approximation percentile) inference supported. BMLMI, pooling inference steps performed via pool() case implements method described Section 3.9.","code":""},{"path":"/articles/stat_specs.html","id":"sec:otherSoftware","dir":"Articles","previous_headings":"","what":"Comparison to other software implementations","title":"rbmi: Statistical Specifications","text":"established software implementation reference-based imputation SAS -called “five macros” James Roger (Roger (2021)). alternative R implementation also currently development R package RefBasedMI (McGrath White (2021)). rbmi several features supported implementations: addition Bayesian MI approach implemented also packages, implementation provides three alternative MI approaches: approximate Bayesian MI, conditional mean imputation combined resampling, bootstrapped MI. rbmi allows usage data collected ICE. example, suppose want adopt treatment policy strategy ICE “treatment discontinuation”. possible implementation strategy use observed outcome data subjects remain study ICE use reference-based imputation case subject drops . implementation, implemented excluding observed post ICE data imputation model assumes MAR missingness including analysis model. knowledge, directly supported implementations. RefBasedMI fits imputation model data treatment group separately implies covariate-treatment group interactions covariates pooled data treatment groups. contrast, Roger’s five macros assume joint model including data randomized groups covariate-treatment interactions covariates allowed. also chose implement joint model use flexible model linear predictor may may include interaction term covariate treatment group. addition, imputation model also allows inclusion time-varying covariates. implementation, grouping subjects purpose imputation model (definition reference group) need correspond assigned treatment groups. provides additional flexibility imputation procedure. clear us whether feature supported Roger’s five macros RefBasedMI. believe R-based implementation modular RefBasedMI facilitate package enhancements. contrast, general causal model introduced . White, Royes, Best (2020) available implementations currently supported .","code":""},{"path":[]},{"path":"/authors.html","id":null,"dir":"","previous_headings":"","what":"Authors","title":"Authors and Citation","text":"Craig Gower-Page. Author, maintainer. Alessandro Noci. Author. Marcel Wolbers. Contributor. Roche. Copyright holder, funder.","code":""},{"path":"/authors.html","id":"citation","dir":"","previous_headings":"","what":"Citation","title":"Authors and Citation","text":"Gower-Page C, Noci (2024). rbmi: Reference Based Multiple Imputation. R package version 1.2.6, https://github.com/insightsengineering/rbmi, https://insightsengineering.github.io/rbmi/.","code":"@Manual{, title = {rbmi: Reference Based Multiple Imputation}, author = {Craig Gower-Page and Alessandro Noci}, year = {2024}, note = {R package version 1.2.6, https://github.com/insightsengineering/rbmi}, url = {https://insightsengineering.github.io/rbmi/}, }"},{"path":[]},{"path":"/index.html","id":"overview","dir":"","previous_headings":"","what":"Overview","title":"Reference Based Multiple Imputation","text":"rbmi R package imputation missing data clinical trials continuous multivariate normal longitudinal outcomes. supports imputation missing random (MAR) assumption, reference-based imputation methods, delta adjustments (required sensitivity analysis tipping point analyses). package implements Bayesian approximate Bayesian multiple imputation combined Rubin’s rules inference, frequentist conditional mean imputation combined (jackknife bootstrap) resampling.","code":""},{"path":"/index.html","id":"installation","dir":"","previous_headings":"","what":"Installation","title":"Reference Based Multiple Imputation","text":"package can installed directly CRAN via:","code":"install.packages(\"rbmi\")"},{"path":"/index.html","id":"usage","dir":"","previous_headings":"","what":"Usage","title":"Reference Based Multiple Imputation","text":"package designed around 4 core functions: draws() - Fits multiple imputation models impute() - Imputes multiple datasets analyse() - Analyses multiple datasets pool() - Pools multiple results single statistic basic usage core functions described quickstart vignette:","code":"vignette(topic = \"quickstart\", package = \"rbmi\")"},{"path":"/index.html","id":"support","dir":"","previous_headings":"","what":"Support","title":"Reference Based Multiple Imputation","text":"help regards using package find bug please create GitHub issue","code":""},{"path":"/reference/QR_decomp.html","id":null,"dir":"Reference","previous_headings":"","what":"QR decomposition — QR_decomp","title":"QR decomposition — QR_decomp","text":"QR decomposition defined Stan user's guide (section 1.2).","code":""},{"path":"/reference/QR_decomp.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"QR decomposition — QR_decomp","text":"","code":"QR_decomp(mat)"},{"path":"/reference/QR_decomp.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"QR decomposition — QR_decomp","text":"mat matrix perform QR decomposition .","code":""},{"path":"/reference/Stack.html","id":null,"dir":"Reference","previous_headings":"","what":"R6 Class for a FIFO stack — Stack","title":"R6 Class for a FIFO stack — Stack","text":"simple stack object offering add / pop functionality","code":""},{"path":"/reference/Stack.html","id":"public-fields","dir":"Reference","previous_headings":"","what":"Public fields","title":"R6 Class for a FIFO stack — Stack","text":"stack list containing current stack","code":""},{"path":[]},{"path":"/reference/Stack.html","id":"public-methods","dir":"Reference","previous_headings":"","what":"Public methods","title":"R6 Class for a FIFO stack — Stack","text":"Stack$add() Stack$pop() Stack$clone()","code":""},{"path":"/reference/Stack.html","id":"method-add-","dir":"Reference","previous_headings":"","what":"Method add()","title":"R6 Class for a FIFO stack — Stack","text":"Adds content end stack (must list)","code":""},{"path":"/reference/Stack.html","id":"usage","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for a FIFO stack — Stack","text":"","code":"Stack$add(x)"},{"path":"/reference/Stack.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for a FIFO stack — Stack","text":"x content add stack","code":""},{"path":"/reference/Stack.html","id":"method-pop-","dir":"Reference","previous_headings":"","what":"Method pop()","title":"R6 Class for a FIFO stack — Stack","text":"Retrieve content stack","code":""},{"path":"/reference/Stack.html","id":"usage-1","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for a FIFO stack — Stack","text":"","code":"Stack$pop(i)"},{"path":"/reference/Stack.html","id":"arguments-1","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for a FIFO stack — Stack","text":"number items retrieve stack. less items left stack just return everything left.","code":""},{"path":"/reference/Stack.html","id":"method-clone-","dir":"Reference","previous_headings":"","what":"Method clone()","title":"R6 Class for a FIFO stack — Stack","text":"objects class cloneable method.","code":""},{"path":"/reference/Stack.html","id":"usage-2","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for a FIFO stack — Stack","text":"","code":"Stack$clone(deep = FALSE)"},{"path":"/reference/Stack.html","id":"arguments-2","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for a FIFO stack — Stack","text":"deep Whether make deep clone.","code":""},{"path":"/reference/add_class.html","id":null,"dir":"Reference","previous_headings":"","what":"Add a class — add_class","title":"Add a class — add_class","text":"Utility function add class object. Adds new class existing classes.","code":""},{"path":"/reference/add_class.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Add a class — add_class","text":"","code":"add_class(x, cls)"},{"path":"/reference/add_class.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Add a class — add_class","text":"x object add class . cls class added.","code":""},{"path":"/reference/adjust_trajectories.html","id":null,"dir":"Reference","previous_headings":"","what":"Adjust trajectories due to the intercurrent event (ICE) — adjust_trajectories","title":"Adjust trajectories due to the intercurrent event (ICE) — adjust_trajectories","text":"Adjust trajectories due intercurrent event (ICE)","code":""},{"path":"/reference/adjust_trajectories.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Adjust trajectories due to the intercurrent event (ICE) — adjust_trajectories","text":"","code":"adjust_trajectories( distr_pars_group, outcome, ids, ind_ice, strategy_fun, distr_pars_ref = NULL )"},{"path":"/reference/adjust_trajectories.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Adjust trajectories due to the intercurrent event (ICE) — adjust_trajectories","text":"distr_pars_group Named list containing simulation parameters multivariate normal distribution assumed given treatment group. contains following elements: mu: Numeric vector indicating mean outcome trajectory. include outcome baseline. sigma Covariance matrix outcome trajectory. outcome Numeric variable specifies longitudinal outcome. ids Factor variable specifies id subject. ind_ice binary variable takes value 1 corresponding outcome affected ICE 0 otherwise. strategy_fun Function implementing trajectories intercurrent event (ICE). Must one getStrategies(). See getStrategies() details. distr_pars_ref Optional. Named list containing simulation parameters reference arm. contains following elements: mu: Numeric vector indicating mean outcome trajectory assuming ICEs. include outcome baseline. sigma Covariance matrix outcome trajectory assuming ICEs.","code":""},{"path":"/reference/adjust_trajectories.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Adjust trajectories due to the intercurrent event (ICE) — adjust_trajectories","text":"numeric vector containing adjusted trajectories.","code":""},{"path":[]},{"path":"/reference/adjust_trajectories_single.html","id":null,"dir":"Reference","previous_headings":"","what":"Adjust trajectory of a subject's outcome due to the intercurrent event (ICE) — adjust_trajectories_single","title":"Adjust trajectory of a subject's outcome due to the intercurrent event (ICE) — adjust_trajectories_single","text":"Adjust trajectory subject's outcome due intercurrent event (ICE)","code":""},{"path":"/reference/adjust_trajectories_single.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Adjust trajectory of a subject's outcome due to the intercurrent event (ICE) — adjust_trajectories_single","text":"","code":"adjust_trajectories_single( distr_pars_group, outcome, strategy_fun, distr_pars_ref = NULL )"},{"path":"/reference/adjust_trajectories_single.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Adjust trajectory of a subject's outcome due to the intercurrent event (ICE) — adjust_trajectories_single","text":"distr_pars_group Named list containing simulation parameters multivariate normal distribution assumed given treatment group. contains following elements: mu: Numeric vector indicating mean outcome trajectory. include outcome baseline. sigma Covariance matrix outcome trajectory. outcome Numeric variable specifies longitudinal outcome. strategy_fun Function implementing trajectories intercurrent event (ICE). Must one getStrategies(). See getStrategies() details. distr_pars_ref Optional. Named list containing simulation parameters reference arm. contains following elements: mu: Numeric vector indicating mean outcome trajectory assuming ICEs. include outcome baseline. sigma Covariance matrix outcome trajectory assuming ICEs.","code":""},{"path":"/reference/adjust_trajectories_single.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Adjust trajectory of a subject's outcome due to the intercurrent event (ICE) — adjust_trajectories_single","text":"numeric vector containing adjusted trajectory single subject.","code":""},{"path":"/reference/adjust_trajectories_single.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Adjust trajectory of a subject's outcome due to the intercurrent event (ICE) — adjust_trajectories_single","text":"outcome specified --post-ICE observations (.e. observations adjusted) set NA.","code":""},{"path":"/reference/analyse.html","id":null,"dir":"Reference","previous_headings":"","what":"Analyse Multiple Imputed Datasets — analyse","title":"Analyse Multiple Imputed Datasets — analyse","text":"function takes multiple imputed datasets (generated impute() function) runs analysis function .","code":""},{"path":"/reference/analyse.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Analyse Multiple Imputed Datasets — analyse","text":"","code":"analyse(imputations, fun = ancova, delta = NULL, ...)"},{"path":"/reference/analyse.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Analyse Multiple Imputed Datasets — analyse","text":"imputations imputations object created impute(). fun analysis function applied imputed dataset. See details. delta data.frame containing delta transformation applied imputed datasets prior running fun. See details. ... Additional arguments passed onto fun.","code":""},{"path":"/reference/analyse.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Analyse Multiple Imputed Datasets — analyse","text":"function works performing following steps: Extract dataset imputations object. Apply delta adjustments specified delta argument. Run analysis function fun dataset. Repeat steps 1-3 across datasets inside imputations object. Collect return analysis results. analysis function fun must take data.frame first argument. options analyse() passed onto fun via .... fun must return named list element list containing single numeric element called est (additionally se df originally specified method_bayes() method_approxbayes()) .e.: Please note vars$subjid column (defined original call draws()) scrambled data.frames provided fun. say contain original subject values hard coding subject ids strictly avoided. default fun ancova() function. Please note function requires vars object, created set_vars(), provided via vars argument e.g. analyse(imputeObj, vars = set_vars(...)). Please see documentation ancova() full details. Please also note theoretical justification conditional mean imputation method (method = method_condmean() draws()) relies fact ANCOVA linear transformation outcomes. Thus care required applying alternative analysis functions setting. delta argument can used specify offsets applied outcome variable imputed datasets prior analysis. typically used sensitivity tipping point analyses. delta dataset must contain columns vars$subjid, vars$visit (specified original call draws()) delta. Essentially data.frame merged onto imputed dataset vars$subjid vars$visit outcome variable modified : Please note order provide maximum flexibility, delta argument can used modify /outcome values including imputed. Care must taken defining offsets. recommend use helper function delta_template() define delta datasets provides utility variables is_missing can used identify exactly visits imputed.","code":"myfun <- function(dat, ...) { mod_1 <- lm(data = dat, outcome ~ group) mod_2 <- lm(data = dat, outcome ~ group + covar) x <- list( trt_1 = list( est = coef(mod_1)[[group]], se = sqrt(vcov(mod_1)[group, group]), df = df.residual(mod_1) ), trt_2 = list( est = coef(mod_2)[[group]], se = sqrt(vcov(mod_2)[group, group]), df = df.residual(mod_2) ) ) return(x) } imputed_data[[vars$outcome]] <- imputed_data[[vars$outcome]] + imputed_data[[\"delta\"]]"},{"path":[]},{"path":"/reference/analyse.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Analyse Multiple Imputed Datasets — analyse","text":"","code":"if (FALSE) { vars <- set_vars( subjid = \"subjid\", visit = \"visit\", outcome = \"outcome\", group = \"group\", covariates = c(\"sex\", \"age\", \"sex*age\") ) analyse( imputations = imputeObj, vars = vars ) deltadf <- data.frame( subjid = c(\"Pt1\", \"Pt1\", \"Pt2\"), visit = c(\"Visit_1\", \"Visit_2\", \"Visit_2\"), delta = c( 5, 9, -10) ) analyse( imputations = imputeObj, delta = deltadf, vars = vars ) }"},{"path":"/reference/ancova.html","id":null,"dir":"Reference","previous_headings":"","what":"Analysis of Covariance — ancova","title":"Analysis of Covariance — ancova","text":"Performs analysis covariance two groups returning estimated \"treatment effect\" (.e. contrast two treatment groups) least square means estimates group.","code":""},{"path":"/reference/ancova.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Analysis of Covariance — ancova","text":"","code":"ancova(data, vars, visits = NULL, weights = c(\"proportional\", \"equal\"))"},{"path":"/reference/ancova.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Analysis of Covariance — ancova","text":"data data.frame containing data used model. vars vars object generated set_vars(). group, visit, outcome covariates elements required. See details. visits optional character vector specifying visits fit ancova model . NULL, separate ancova model fit outcomes visit (determined unique(data[[vars$visit]])). See details. weights Character, either \"proportional\" (default) \"equal\". Specifies weighting strategy used categorical covariates calculating lsmeans. See details.","code":""},{"path":"/reference/ancova.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Analysis of Covariance — ancova","text":"function works follows: Select first value visits. Subset data observations occurred visit. Fit linear model vars$outcome ~ vars$group + vars$covariates. Extract \"treatment effect\" & least square means treatment group. Repeat points 2-3 values visits. value visits provided set unique(data[[vars$visit]]). order meet formatting standards set analyse() results collapsed single list suffixed visit name, e.g.: Please note \"ref\" refers first factor level vars$group necessarily coincide control arm. Analogously, \"alt\" refers second factor level vars$group. \"trt\" refers model contrast translating mean difference second level first level. want include interaction terms model can done providing covariates argument set_vars() e.g. set_vars(covariates = c(\"sex*age\")).","code":"list( trt_visit_1 = list(est = ...), lsm_ref_visit_1 = list(est = ...), lsm_alt_visit_1 = list(est = ...), trt_visit_2 = list(est = ...), lsm_ref_visit_2 = list(est = ...), lsm_alt_visit_2 = list(est = ...), ... )"},{"path":"/reference/ancova.html","id":"weighting","dir":"Reference","previous_headings":"","what":"Weighting","title":"Analysis of Covariance — ancova","text":"\"proportional\" default scheme used. equivalent standardization, .e. lsmeans group equal predicted mean outcome ancova model group based baseline characteristics subjects regardless assigned group. alternative weighting scheme, \"equal\", creates hypothetical patients expanding combinations models categorical covariates. lsmeans calculated average predicted mean outcome hypothetical patients assuming come group turn. short: \"proportional\" weights categorical covariates based upon frequency occurrence data. \"equal\" weights categorical covariates equally across theoretical combinations.","code":""},{"path":[]},{"path":"/reference/ancova_single.html","id":null,"dir":"Reference","previous_headings":"","what":"Implements an Analysis of Covariance (ANCOVA) — ancova_single","title":"Implements an Analysis of Covariance (ANCOVA) — ancova_single","text":"Performance analysis covariance. See ancova() full details.","code":""},{"path":"/reference/ancova_single.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Implements an Analysis of Covariance (ANCOVA) — ancova_single","text":"","code":"ancova_single( data, outcome, group, covariates, weights = c(\"proportional\", \"equal\") )"},{"path":"/reference/ancova_single.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Implements an Analysis of Covariance (ANCOVA) — ancova_single","text":"data data.frame containing data required model. outcome Character, name outcome variable data. group Character, name group variable data. covariates Character vector containing name additional covariates included model well interaction terms. weights Character, specifies whether use \"proportional\" \"equal\" weighting categorical covariate combination calculating lsmeans.","code":""},{"path":"/reference/ancova_single.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Implements an Analysis of Covariance (ANCOVA) — ancova_single","text":"group must factor variable 2 levels. outcome must continuous numeric variable.","code":""},{"path":[]},{"path":"/reference/ancova_single.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Implements an Analysis of Covariance (ANCOVA) — ancova_single","text":"","code":"if (FALSE) { iris2 <- iris[ iris$Species %in% c(\"versicolor\", \"virginica\"), ] iris2$Species <- factor(iris2$Species) ancova_single(iris2, \"Sepal.Length\", \"Species\", c(\"Petal.Length * Petal.Width\")) }"},{"path":"/reference/antidepressant_data.html","id":null,"dir":"Reference","previous_headings":"","what":"Antidepressant trial data — antidepressant_data","title":"Antidepressant trial data — antidepressant_data","text":"dataset containing data publicly available example data set antidepressant clinical trial. dataset available website Drug Information Association Scientific Working Group Estimands Missing Data. per website, original data antidepressant clinical trial four treatments; two doses experimental medication, positive control, placebo published Goldstein et al (2004). mask real data, week 8 observations removed two arms created: original placebo arm \"drug arm\" created randomly selecting patients three non-placebo arms.","code":""},{"path":"/reference/antidepressant_data.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Antidepressant trial data — antidepressant_data","text":"","code":"antidepressant_data"},{"path":"/reference/antidepressant_data.html","id":"format","dir":"Reference","previous_headings":"","what":"Format","title":"Antidepressant trial data — antidepressant_data","text":"data.frame 608 rows 11 variables: PATIENT: patients IDs. HAMATOTL: total score Hamilton Anxiety Rating Scale. PGIIMP: patient's Global Impression Improvement Rating Scale. RELDAYS: number days visit baseline. VISIT: post-baseline visit. levels 4,5,6,7. THERAPY: treatment group variable. equal PLACEBO observations placebo arm, DRUG observations active arm. GENDER: patient's gender. POOLINV: pooled investigator. BASVAL: baseline outcome value. HAMDTL17: Hamilton 17-item rating scale value. CHANGE: change baseline Hamilton 17-item rating scale.","code":""},{"path":"/reference/antidepressant_data.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Antidepressant trial data — antidepressant_data","text":"relevant endpoint Hamilton 17-item rating scale depression (HAMD17) baseline weeks 1, 2, 4, 6 assessments included. Study drug discontinuation occurred 24% subjects active drug 26% placebo. data study drug discontinuation missing single additional intermittent missing observation.","code":""},{"path":"/reference/antidepressant_data.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Antidepressant trial data — antidepressant_data","text":"Goldstein, Lu, Detke, Wiltse, Mallinckrodt, Demitrack. Duloxetine treatment depression: double-blind placebo-controlled comparison paroxetine. J Clin Psychopharmacol 2004;24: 389-399.","code":""},{"path":"/reference/apply_delta.html","id":null,"dir":"Reference","previous_headings":"","what":"Applies delta adjustment — apply_delta","title":"Applies delta adjustment — apply_delta","text":"Takes delta dataset adjusts outcome variable adding corresponding delta.","code":""},{"path":"/reference/apply_delta.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Applies delta adjustment — apply_delta","text":"","code":"apply_delta(data, delta = NULL, group = NULL, outcome = NULL)"},{"path":"/reference/apply_delta.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Applies delta adjustment — apply_delta","text":"data data.frame outcome column adjusted. delta data.frame (must contain column called delta). group character vector variables data delta used merge 2 data.frames together . outcome character, name outcome variable data.","code":""},{"path":"/reference/as_analysis.html","id":null,"dir":"Reference","previous_headings":"","what":"Construct an analysis object — as_analysis","title":"Construct an analysis object — as_analysis","text":"Creates analysis object ensuring components correctly defined.","code":""},{"path":"/reference/as_analysis.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Construct an analysis object — as_analysis","text":"","code":"as_analysis(results, method, delta = NULL, fun = NULL, fun_name = NULL)"},{"path":"/reference/as_analysis.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Construct an analysis object — as_analysis","text":"results list lists contain analysis results imputation See analyse() details object look like. method method object specified draws(). delta delta dataset used. See analyse() details specified. fun analysis function used. fun_name character name analysis function (used printing) purposes.","code":""},{"path":"/reference/as_ascii_table.html","id":null,"dir":"Reference","previous_headings":"","what":"as_ascii_table — as_ascii_table","title":"as_ascii_table — as_ascii_table","text":"function takes data.frame attempts convert simple ascii format suitable printing screen assumed variable values .character() method order cast character.","code":""},{"path":"/reference/as_ascii_table.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"as_ascii_table — as_ascii_table","text":"","code":"as_ascii_table(dat, line_prefix = \" \", pcol = NULL)"},{"path":"/reference/as_ascii_table.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"as_ascii_table — as_ascii_table","text":"dat Input dataset convert ascii table line_prefix Symbols prefix infront every line table pcol name column handled p-value. Sets value <0.001 value 0 rounding","code":""},{"path":"/reference/as_class.html","id":null,"dir":"Reference","previous_headings":"","what":"Set Class — as_class","title":"Set Class — as_class","text":"Utility function set objects class.","code":""},{"path":"/reference/as_class.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Set Class — as_class","text":"","code":"as_class(x, cls)"},{"path":"/reference/as_class.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Set Class — as_class","text":"x object set class . cls class set.","code":""},{"path":"/reference/as_cropped_char.html","id":null,"dir":"Reference","previous_headings":"","what":"as_cropped_char — as_cropped_char","title":"as_cropped_char — as_cropped_char","text":"Makes character string x chars Reduce x char string ...","code":""},{"path":"/reference/as_cropped_char.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"as_cropped_char — as_cropped_char","text":"","code":"as_cropped_char(inval, crop_at = 30, ndp = 3)"},{"path":"/reference/as_cropped_char.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"as_cropped_char — as_cropped_char","text":"inval single element value crop_at character limit ndp Number decimal places display","code":""},{"path":"/reference/as_dataframe.html","id":null,"dir":"Reference","previous_headings":"","what":"Convert object to dataframe — as_dataframe","title":"Convert object to dataframe — as_dataframe","text":"Convert object dataframe","code":""},{"path":"/reference/as_dataframe.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Convert object to dataframe — as_dataframe","text":"","code":"as_dataframe(x)"},{"path":"/reference/as_dataframe.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Convert object to dataframe — as_dataframe","text":"x data.frame like object Utility function convert \"data.frame-like\" object actual data.frame avoid issues inconsistency methods ( [() dplyr's grouped dataframes)","code":""},{"path":"/reference/as_draws.html","id":null,"dir":"Reference","previous_headings":"","what":"Creates a draws object — as_draws","title":"Creates a draws object — as_draws","text":"Creates draws object final output call draws().","code":""},{"path":"/reference/as_draws.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Creates a draws object — as_draws","text":"","code":"as_draws(method, samples, data, formula, n_failures = NULL, fit = NULL)"},{"path":"/reference/as_draws.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Creates a draws object — as_draws","text":"method method object generated either method_bayes(), method_approxbayes(), method_condmean() method_bmlmi(). samples list sample_single objects. See sample_single(). data R6 longdata object containing relevant input data information. formula Fixed effects formula object used model specification. n_failures Absolute number failures model fit. fit method_bayes() chosen, returns MCMC Stan fit object. Otherwise NULL.","code":""},{"path":"/reference/as_draws.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Creates a draws object — as_draws","text":"draws object named list containing following: data: R6 longdata object containing relevant input data information. method: method object generated either method_bayes(), method_approxbayes() method_condmean(). samples: list containing estimated parameters interest. element samples named list containing following: ids: vector characters containing ids subjects included original dataset. beta: numeric vector estimated regression coefficients. sigma: list estimated covariance matrices (one level vars$group). theta: numeric vector transformed covariances. failed: Logical. TRUE model fit failed. ids_samp: vector characters containing ids subjects included given sample. fit: method_bayes() chosen, returns MCMC Stan fit object. Otherwise NULL. n_failures: absolute number failures model fit. Relevant method_condmean(type = \"bootstrap\"), method_approxbayes() method_bmlmi(). formula: fixed effects formula object used model specification.","code":""},{"path":"/reference/as_imputation.html","id":null,"dir":"Reference","previous_headings":"","what":"Create an imputation object — as_imputation","title":"Create an imputation object — as_imputation","text":"function creates object returned impute(). Essentially glorified wrapper around list() ensuring required elements set class added expected.","code":""},{"path":"/reference/as_imputation.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create an imputation object — as_imputation","text":"","code":"as_imputation(imputations, data, method, references)"},{"path":"/reference/as_imputation.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create an imputation object — as_imputation","text":"imputations list imputations_list's created imputation_df() data longdata object created longDataConstructor() method method object created method_condmean(), method_bayes() method_approxbayes() references named vector. Identifies references used generating imputed values. form c(\"Group\" = \"Reference\", \"Group\" = \"Reference\").","code":""},{"path":"/reference/as_indices.html","id":null,"dir":"Reference","previous_headings":"","what":"Convert indicator to index — as_indices","title":"Convert indicator to index — as_indices","text":"Converts string 0's 1's index positions 1's padding results 0's length","code":""},{"path":"/reference/as_indices.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Convert indicator to index — as_indices","text":"","code":"as_indices(x)"},{"path":"/reference/as_indices.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Convert indicator to index — as_indices","text":"x character vector whose values either \"0\" \"1\". elements vector must length","code":""},{"path":"/reference/as_indices.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Convert indicator to index — as_indices","text":".e.","code":"patmap(c(\"1101\", \"0001\")) -> list(c(1,2,4,999), c(4,999, 999, 999))"},{"path":"/reference/as_mmrm_df.html","id":null,"dir":"Reference","previous_headings":"","what":"Creates a ","title":"Creates a ","text":"Converts design matrix + key variables common format particular function following: Renames covariates V1, V2, etc avoid issues special characters variable names Ensures key variables right type Inserts outcome, visit subjid variables data.frame naming outcome, visit subjid provided also insert group variable data.frame named group","code":""},{"path":"/reference/as_mmrm_df.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Creates a ","text":"","code":"as_mmrm_df(designmat, outcome, visit, subjid, group = NULL)"},{"path":"/reference/as_mmrm_df.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Creates a ","text":"designmat data.frame matrix containing covariates use MMRM model. Dummy variables must already expanded , .e. via stats::model.matrix(). contain missing values outcome numeric vector. outcome value regressed MMRM model. visit character / factor vector. Indicates visit outcome value occurred . subjid character / factor vector. subject identifier used link separate visits belong subject. group character / factor vector. Indicates treatment group patient belongs .","code":""},{"path":"/reference/as_mmrm_formula.html","id":null,"dir":"Reference","previous_headings":"","what":"Create MMRM formula — as_mmrm_formula","title":"Create MMRM formula — as_mmrm_formula","text":"Derives MMRM model formula structure mmrm_df. returns formula object form:","code":""},{"path":"/reference/as_mmrm_formula.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create MMRM formula — as_mmrm_formula","text":"","code":"as_mmrm_formula(mmrm_df, cov_struct)"},{"path":"/reference/as_mmrm_formula.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create MMRM formula — as_mmrm_formula","text":"mmrm_df mmrm data.frame created as_mmrm_df() cov_struct Character - covariance structure used, must one \"us\", \"toep\", \"cs\", \"ar1\"","code":""},{"path":"/reference/as_mmrm_formula.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Create MMRM formula — as_mmrm_formula","text":"","code":"outcome ~ 0 + V1 + V2 + V4 + ... + us(visit | group / subjid)"},{"path":"/reference/as_model_df.html","id":null,"dir":"Reference","previous_headings":"","what":"Expand data.frame into a design matrix — as_model_df","title":"Expand data.frame into a design matrix — as_model_df","text":"Expands data.frame using formula create design matrix. Key details always place outcome variable first column return object.","code":""},{"path":"/reference/as_model_df.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Expand data.frame into a design matrix — as_model_df","text":"","code":"as_model_df(dat, frm)"},{"path":"/reference/as_model_df.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Expand data.frame into a design matrix — as_model_df","text":"dat data.frame frm formula","code":""},{"path":"/reference/as_model_df.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Expand data.frame into a design matrix — as_model_df","text":"outcome column may contain NA's none variables listed formula contain missing values","code":""},{"path":"/reference/as_simple_formula.html","id":null,"dir":"Reference","previous_headings":"","what":"Creates a simple formula object from a string — as_simple_formula","title":"Creates a simple formula object from a string — as_simple_formula","text":"Converts string list variables formula object","code":""},{"path":"/reference/as_simple_formula.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Creates a simple formula object from a string — as_simple_formula","text":"","code":"as_simple_formula(outcome, covars)"},{"path":"/reference/as_simple_formula.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Creates a simple formula object from a string — as_simple_formula","text":"outcome character (length 1 vector). Name outcome variable covars character (vector). Name covariates","code":""},{"path":"/reference/as_simple_formula.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Creates a simple formula object from a string — as_simple_formula","text":"formula","code":""},{"path":"/reference/as_stan_array.html","id":null,"dir":"Reference","previous_headings":"","what":"As array — as_stan_array","title":"As array — as_stan_array","text":"Converts numeric value length 1 1 dimension array. avoid type errors thrown stan length 1 numeric vectors provided R stan::vector inputs","code":""},{"path":"/reference/as_stan_array.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"As array — as_stan_array","text":"","code":"as_stan_array(x)"},{"path":"/reference/as_stan_array.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"As array — as_stan_array","text":"x numeric vector","code":""},{"path":"/reference/as_strata.html","id":null,"dir":"Reference","previous_headings":"","what":"Create vector of Stratas — as_strata","title":"Create vector of Stratas — as_strata","text":"Collapse multiple categorical variables distinct unique categories. e.g. return","code":"as_strata(c(1,1,2,2,2,1), c(5,6,5,5,6,5)) c(1,2,3,3,4,1)"},{"path":"/reference/as_strata.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create vector of Stratas — as_strata","text":"","code":"as_strata(...)"},{"path":"/reference/as_strata.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create vector of Stratas — as_strata","text":"... numeric/character/factor vectors length","code":""},{"path":"/reference/as_strata.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Create vector of Stratas — as_strata","text":"","code":"if (FALSE) { as_strata(c(1,1,2,2,2,1), c(5,6,5,5,6,5)) }"},{"path":"/reference/assert_variables_exist.html","id":null,"dir":"Reference","previous_headings":"","what":"Assert that all variables exist within a dataset — assert_variables_exist","title":"Assert that all variables exist within a dataset — assert_variables_exist","text":"Performs assertion check ensure vector variable exists within data.frame expected.","code":""},{"path":"/reference/assert_variables_exist.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Assert that all variables exist within a dataset — assert_variables_exist","text":"","code":"assert_variables_exist(data, vars)"},{"path":"/reference/assert_variables_exist.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Assert that all variables exist within a dataset — assert_variables_exist","text":"data data.frame vars character vector variable names","code":""},{"path":"/reference/char2fct.html","id":null,"dir":"Reference","previous_headings":"","what":"Convert character variables to factor — char2fct","title":"Convert character variables to factor — char2fct","text":"Provided vector variable names function converts character variables factors. affect numeric existing factor variables","code":""},{"path":"/reference/char2fct.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Convert character variables to factor — char2fct","text":"","code":"char2fct(data, vars = NULL)"},{"path":"/reference/char2fct.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Convert character variables to factor — char2fct","text":"data data.frame vars character vector variables data","code":""},{"path":"/reference/check_ESS.html","id":null,"dir":"Reference","previous_headings":"","what":"Diagnostics of the MCMC based on ESS — check_ESS","title":"Diagnostics of the MCMC based on ESS — check_ESS","text":"Check quality MCMC draws posterior distribution checking whether relative ESS sufficiently large.","code":""},{"path":"/reference/check_ESS.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Diagnostics of the MCMC based on ESS — check_ESS","text":"","code":"check_ESS(stan_fit, n_draws, threshold_lowESS = 0.4)"},{"path":"/reference/check_ESS.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Diagnostics of the MCMC based on ESS — check_ESS","text":"stan_fit stanfit object. n_draws Number MCMC draws. threshold_lowESS number [0,1] indicating minimum acceptable value relative ESS. See details.","code":""},{"path":"/reference/check_ESS.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Diagnostics of the MCMC based on ESS — check_ESS","text":"warning message case detected problems.","code":""},{"path":"/reference/check_ESS.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Diagnostics of the MCMC based on ESS — check_ESS","text":"check_ESS() works follows: Extract ESS stan_fit parameter model. Compute relative ESS (.e. ESS divided number draws). Check whether parameter ESS lower threshold. least one parameter relative ESS threshold, warning thrown.","code":""},{"path":"/reference/check_hmc_diagn.html","id":null,"dir":"Reference","previous_headings":"","what":"Diagnostics of the MCMC based on HMC-related measures. — check_hmc_diagn","title":"Diagnostics of the MCMC based on HMC-related measures. — check_hmc_diagn","text":"Check : divergent iterations. Bayesian Fraction Missing Information (BFMI) sufficiently low. number iterations saturated max treedepth zero. Please see rstan::check_hmc_diagnostics() details.","code":""},{"path":"/reference/check_hmc_diagn.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Diagnostics of the MCMC based on HMC-related measures. — check_hmc_diagn","text":"","code":"check_hmc_diagn(stan_fit)"},{"path":"/reference/check_hmc_diagn.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Diagnostics of the MCMC based on HMC-related measures. — check_hmc_diagn","text":"stan_fit stanfit object.","code":""},{"path":"/reference/check_hmc_diagn.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Diagnostics of the MCMC based on HMC-related measures. — check_hmc_diagn","text":"warning message case detected problems.","code":""},{"path":"/reference/check_mcmc.html","id":null,"dir":"Reference","previous_headings":"","what":"Diagnostics of the MCMC — check_mcmc","title":"Diagnostics of the MCMC — check_mcmc","text":"Diagnostics MCMC","code":""},{"path":"/reference/check_mcmc.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Diagnostics of the MCMC — check_mcmc","text":"","code":"check_mcmc(stan_fit, n_draws, threshold_lowESS = 0.4)"},{"path":"/reference/check_mcmc.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Diagnostics of the MCMC — check_mcmc","text":"stan_fit stanfit object. n_draws Number MCMC draws. threshold_lowESS number [0,1] indicating minimum acceptable value relative ESS. See details.","code":""},{"path":"/reference/check_mcmc.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Diagnostics of the MCMC — check_mcmc","text":"warning message case detected problems.","code":""},{"path":"/reference/check_mcmc.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Diagnostics of the MCMC — check_mcmc","text":"Performs checks quality MCMC. See check_ESS() check_hmc_diagn() details.","code":""},{"path":"/reference/compute_sigma.html","id":null,"dir":"Reference","previous_headings":"","what":"Compute covariance matrix for some reference-based methods (JR, CIR) — compute_sigma","title":"Compute covariance matrix for some reference-based methods (JR, CIR) — compute_sigma","text":"Adapt covariance matrix reference-based methods. Used Copy Increments Reference (CIR) Jump Reference (JTR) methods, adapt covariance matrix different pre-deviation post deviation covariance structures. See Carpenter et al. (2013)","code":""},{"path":"/reference/compute_sigma.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Compute covariance matrix for some reference-based methods (JR, CIR) — compute_sigma","text":"","code":"compute_sigma(sigma_group, sigma_ref, index_mar)"},{"path":"/reference/compute_sigma.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Compute covariance matrix for some reference-based methods (JR, CIR) — compute_sigma","text":"sigma_group covariance matrix dimensions equal index_mar subjects original group sigma_ref covariance matrix dimensions equal index_mar subjects reference group index_mar logical vector indicating visits meet MAR assumption subject. .e. identifies observations non-MAR intercurrent event (ICE).","code":""},{"path":"/reference/compute_sigma.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Compute covariance matrix for some reference-based methods (JR, CIR) — compute_sigma","text":"Carpenter, James R., James H. Roger, Michael G. Kenward. \"Analysis longitudinal trials protocol deviation: framework relevant, accessible assumptions, inference via multiple imputation.\" Journal Biopharmaceutical statistics 23.6 (2013): 1352-1371.","code":""},{"path":"/reference/convert_to_imputation_list_df.html","id":null,"dir":"Reference","previous_headings":"","what":"Convert list of imputation_list_single() objects to an imputation_list_df() object\n(i.e. a list of imputation_df() objects's) — convert_to_imputation_list_df","title":"Convert list of imputation_list_single() objects to an imputation_list_df() object\n(i.e. a list of imputation_df() objects's) — convert_to_imputation_list_df","text":"Convert list imputation_list_single() objects imputation_list_df() object (.e. list imputation_df() objects's)","code":""},{"path":"/reference/convert_to_imputation_list_df.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Convert list of imputation_list_single() objects to an imputation_list_df() object\n(i.e. a list of imputation_df() objects's) — convert_to_imputation_list_df","text":"","code":"convert_to_imputation_list_df(imputes, sample_ids)"},{"path":"/reference/convert_to_imputation_list_df.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Convert list of imputation_list_single() objects to an imputation_list_df() object\n(i.e. a list of imputation_df() objects's) — convert_to_imputation_list_df","text":"imputes list imputation_list_single() objects sample_ids list 1 element per required imputation_df. element must contain vector \"ID\"'s correspond imputation_single() ID's required dataset. total number ID's must equal total number rows within imputes$imputations accommodate method_bmlmi() impute_data_individual() function returns list imputation_list_single() objects 1 object per subject. imputation_list_single() stores subjects imputations matrix columns matrix correspond D method_bmlmi(). Note methods (.e. methods_*()) special case D = 1. number rows matrix varies subject equal number times patient selected imputation (non-conditional mean methods 1 per subject per imputed dataset). function best illustrated example: convert_to_imputation_df(imputes, sample_ids) result : Note different repetitions (.e. value set D) grouped together sequentially.","code":"imputes = list( imputation_list_single( id = \"Tom\", imputations = matrix( imputation_single_t_1_1, imputation_single_t_1_2, imputation_single_t_2_1, imputation_single_t_2_2, imputation_single_t_3_1, imputation_single_t_3_2 ) ), imputation_list_single( id = \"Tom\", imputations = matrix( imputation_single_h_1_1, imputation_single_h_1_2, ) ) ) sample_ids <- list( c(\"Tom\", \"Harry\", \"Tom\"), c(\"Tom\") ) imputation_list_df( imputation_df( imputation_single_t_1_1, imputation_single_h_1_1, imputation_single_t_2_1 ), imputation_df( imputation_single_t_1_2, imputation_single_h_1_2, imputation_single_t_2_2 ), imputation_df( imputation_single_t_3_1 ), imputation_df( imputation_single_t_3_2 ) )"},{"path":"/reference/d_lagscale.html","id":null,"dir":"Reference","previous_headings":"","what":"Calculate delta from a lagged scale coefficient — d_lagscale","title":"Calculate delta from a lagged scale coefficient — d_lagscale","text":"Calculates delta value based upon baseline delta value post ICE scaling coefficient.","code":""},{"path":"/reference/d_lagscale.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Calculate delta from a lagged scale coefficient — d_lagscale","text":"","code":"d_lagscale(delta, dlag, is_post_ice)"},{"path":"/reference/d_lagscale.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Calculate delta from a lagged scale coefficient — d_lagscale","text":"delta numeric vector. Determines baseline amount delta applied visit. dlag numeric vector. Determines scaling applied delta based upon visit ICE occurred . Must length delta. is_post_ice logical vector. Indicates whether visit \"post-ICE\" .","code":""},{"path":"/reference/d_lagscale.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Calculate delta from a lagged scale coefficient — d_lagscale","text":"See delta_template() full details calculation performed.","code":""},{"path":"/reference/delta_template.html","id":null,"dir":"Reference","previous_headings":"","what":"Create a delta data.frame template — delta_template","title":"Create a delta data.frame template — delta_template","text":"Creates data.frame format required analyse() use applying delta adjustment.","code":""},{"path":"/reference/delta_template.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create a delta data.frame template — delta_template","text":"","code":"delta_template(imputations, delta = NULL, dlag = NULL, missing_only = TRUE)"},{"path":"/reference/delta_template.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create a delta data.frame template — delta_template","text":"imputations imputation object created impute(). delta NULL numeric vector. Determines baseline amount delta applied visit. See details. numeric vector must length number unique visits original dataset. dlag NULL numeric vector. Determines scaling applied delta based upon visit ICE occurred . See details. numeric vector must length number unique visits original dataset. missing_only Logical, TRUE non-missing post-ICE data delta value 0 assigned. Note calculation (described details section) performed first overwritten 0's end (.e. delta values missing post-ICE visits stay regardless option).","code":""},{"path":"/reference/delta_template.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Create a delta data.frame template — delta_template","text":"apply delta adjustment analyse() function expects delta data.frame 3 variables: vars$subjid, vars$visit delta (vars object supplied original call draws() created set_vars() function). function return data.frame aforementioned variables one row per subject per visit. delta argument function NULL delta column returned data.frame 0 observations. delta argument NULL delta calculated separately subject accumulative sum delta multiplied scaling coefficient dlag based upon many visits subject's intercurrent event (ICE) visit question . best illustrated example: Let delta = c(5,6,7,8) dlag=c(1,2,3,4) (.e. assuming 4 visits) lets say subject ICE visit 2. calculation follows: say subject delta offset 0 applied visit-1, 6 visit-2, 20 visit-3 44 visit-4. comparison, lets say subject instead ICE visit 3, calculation follows: terms practical usage, lets say wanted delta 5 used post ICE visits regardless proximity ICE visit. can achieved setting delta = c(5,5,5,5) dlag = c(1,0,0,0). example lets say subject ICE visit-1, calculation follows: Another way using arguments set delta difference time visits dlag amount delta per unit time. example lets say visit weeks 1, 5, 6 & 9 want delta 3 applied week ICE. can achieved setting delta = c(0,4,1,3) (difference weeks visit) dlag = c(3, 3, 3, 3). example lets say subject ICE week-5 (.e. visit-2) calculation : .e. week-6 (1 week ICE) delta 3 week-9 (4 weeks ICE) delta 12. Please note function also returns several utility variables user can create custom logic defining delta set . additional variables include: is_mar - observation missing regarded MAR? variable set FALSE observations occurred non-MAR ICE, otherwise set TRUE. is_missing - outcome variable observation missing. is_post_ice - observation occur patient's ICE defined data_ice dataset supplied draws(). strategy - imputation strategy assigned subject. design implementation function largely based upon functionality implemented called \"five marcos\" James Roger. See Roger (2021).","code":"v1 v2 v3 v4 -------------- 5 6 7 8 # delta assigned to each visit 0 1 2 3 # lagged scaling starting from the first visit after the subjects ICE -------------- 0 6 14 24 # delta * lagged scaling -------------- 0 6 20 44 # accumulative sum of delta to be applied to each visit v1 v2 v3 v4 -------------- 5 6 7 8 # delta assigned to each visit 0 0 1 2 # lagged scaling starting from the first visit after the subjects ICE -------------- 0 0 7 16 # delta * lagged scaling -------------- 0 0 7 23 # accumulative sum of delta to be applied to each visit v1 v2 v3 v4 -------------- 5 5 5 5 # delta assigned to each visit 1 0 0 0 # lagged scaling starting from the first visit after the subjects ICE -------------- 5 0 0 0 # delta * lagged scaling -------------- 5 5 5 5 # accumulative sum of delta to be applied to each visit v1 v2 v3 v4 -------------- 0 4 1 3 # delta assigned to each visit 0 0 3 3 # lagged scaling starting from the first visit after the subjects ICE -------------- 0 0 3 9 # delta * lagged scaling -------------- 0 0 3 12 # accumulative sum of delta to be applied to each visit"},{"path":"/reference/delta_template.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Create a delta data.frame template — delta_template","text":"Roger, James. Reference-based mi via multivariate normal rm (“five macros” miwithd), 2021. URL https://www.lshtm.ac.uk/research/centres-projects-groups/missing-data#dia-missing-data.","code":""},{"path":[]},{"path":"/reference/delta_template.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Create a delta data.frame template — delta_template","text":"","code":"if (FALSE) { delta_template(imputeObj) delta_template(imputeObj, delta = c(5,6,7,8), dlag = c(1,2,3,4)) }"},{"path":"/reference/do_not_run.html","id":null,"dir":"Reference","previous_headings":"","what":"Do not run this function — do_not_run","title":"Do not run this function — do_not_run","text":"function exists suppress false positive R CMD Check unused libraries","code":""},{"path":"/reference/do_not_run.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Do not run this function — do_not_run","text":"","code":"do_not_run()"},{"path":"/reference/do_not_run.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Do not run this function — do_not_run","text":"rstantools RcppParallel required used installation time. case RcppParallel used src/Makevars file created fly installation rstantools. rstantools used configure file.","code":""},{"path":"/reference/draws.html","id":null,"dir":"Reference","previous_headings":"","what":"Fit the base imputation model and get parameter estimates — draws","title":"Fit the base imputation model and get parameter estimates — draws","text":"draws fits base imputation model observed outcome data according given multiple imputation methodology. According user's method specification, returns either draws posterior distribution model parameters required Bayesian multiple imputation frequentist parameter estimates original data bootstrapped leave-one-datasets required conditional mean imputation. purpose imputation model estimate model parameters absence intercurrent events (ICEs) handled using reference-based imputation methods. reason, observed outcome data ICEs, reference-based imputation methods specified, removed considered missing purpose estimating imputation model, purpose . imputation model mixed model repeated measures (MMRM) valid missing--random (MAR) assumption. can fit using maximum likelihood (ML) restricted ML (REML) estimation, Bayesian approach, approximate Bayesian approach according user's method specification. ML/REML approaches approximate Bayesian approach support several possible covariance structures, Bayesian approach based MCMC sampling supports unstructured covariance structure. case covariance matrix can assumed different across group.","code":""},{"path":"/reference/draws.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Fit the base imputation model and get parameter estimates — draws","text":"","code":"draws(data, data_ice = NULL, vars, method, ncores = 1, quiet = FALSE) # S3 method for approxbayes draws(data, data_ice = NULL, vars, method, ncores = 1, quiet = FALSE) # S3 method for condmean draws(data, data_ice = NULL, vars, method, ncores = 1, quiet = FALSE) # S3 method for bmlmi draws(data, data_ice = NULL, vars, method, ncores = 1, quiet = FALSE) # S3 method for bayes draws(data, data_ice = NULL, vars, method, ncores = 1, quiet = FALSE)"},{"path":"/reference/draws.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Fit the base imputation model and get parameter estimates — draws","text":"data data.frame containing data used model. See details. data_ice data.frame specifies information related ICEs imputation strategies. See details. vars vars object generated set_vars(). See details. method method object generated either method_bayes(), method_approxbayes(), method_condmean() method_bmlmi(). specifies multiple imputation methodology used. See details. ncores single numeric specifying number cores use creating draws object. Note parameter ignored method_bayes() (Default = 1). quiet Logical, TRUE suppress printing progress information printed console.","code":""},{"path":"/reference/draws.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Fit the base imputation model and get parameter estimates — draws","text":"draws object named list containing following: data: R6 longdata object containing relevant input data information. method: method object generated either method_bayes(), method_approxbayes() method_condmean(). samples: list containing estimated parameters interest. element samples named list containing following: ids: vector characters containing ids subjects included original dataset. beta: numeric vector estimated regression coefficients. sigma: list estimated covariance matrices (one level vars$group). theta: numeric vector transformed covariances. failed: Logical. TRUE model fit failed. ids_samp: vector characters containing ids subjects included given sample. fit: method_bayes() chosen, returns MCMC Stan fit object. Otherwise NULL. n_failures: absolute number failures model fit. Relevant method_condmean(type = \"bootstrap\"), method_approxbayes() method_bmlmi(). formula: fixed effects formula object used model specification.","code":""},{"path":"/reference/draws.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Fit the base imputation model and get parameter estimates — draws","text":"draws performs first step multiple imputation (MI) procedure: fitting base imputation model. goal estimate parameters interest needed imputation phase (.e. regression coefficients covariance matrices MMRM model). function distinguishes following methods: Bayesian MI based MCMC sampling: draws returns draws posterior distribution parameters using Bayesian approach based MCMC sampling. method can specified using method = method_bayes(). Approximate Bayesian MI based bootstrapping: draws returns draws posterior distribution parameters using approximate Bayesian approach, sampling posterior distribution simulated fitting MMRM model bootstrap samples original dataset. method can specified using method = method_approxbayes()]. Conditional mean imputation bootstrap re-sampling: draws returns MMRM parameter estimates original dataset n_samples bootstrap samples. method can specified using method = method_condmean() argument type = \"bootstrap\". Conditional mean imputation jackknife re-sampling: draws returns MMRM parameter estimates original dataset leave-one-subject-sample. method can specified using method = method_condmean() argument type = \"jackknife\". Bootstrapped Maximum Likelihood MI: draws returns MMRM parameter estimates given number bootstrap samples needed perform random imputations bootstrapped samples. method can specified using method = method_bmlmi(). Bayesian MI based MCMC sampling proposed Carpenter, Roger, Kenward (2013) first introduced reference-based imputation methods. Approximate Bayesian MI discussed Little Rubin (2002). Conditional mean imputation methods discussed Wolbers et al (2022). Bootstrapped Maximum Likelihood MI described Von Hippel & Bartlett (2021). argument data contains longitudinal data. must least following variables: subjid: factor vector containing subject ids. visit: factor vector containing visit outcome observed . group: factor vector containing group subject belongs . outcome: numeric vector containing outcome variable. might contain missing values. Additional baseline time-varying covariates must included data. data must one row per visit per subject. means incomplete outcome data must set NA instead related row missing. Missing values covariates allowed. data incomplete expand_locf() helper function can used insert missing rows using Last Observation Carried Forward (LOCF) imputation impute covariates values. Note LOCF generally principled imputation method used appropriate specific covariate. Please note special provisioning baseline outcome values. want baseline observations included model part response variable removed advance outcome variable data. time want include baseline outcome covariate model, included separate column data (covariate). Character covariates explicitly cast factors. use custom analysis function requires specific reference levels character covariates (example computation least square means computation) advised manually cast character covariates factor advance running draws(). argument data_ice contains information occurrence ICEs. data.frame 3 columns: Subject ID: character vector containing ids subjects experienced ICE. column must named specified vars$subjid. Visit: character vector containing first visit occurrence ICE (.e. first visit affected ICE). visits must equal one levels data[[vars$visit]]. multiple ICEs happen subject, first non-MAR visit used. column must named specified vars$visit. Strategy: character vector specifying imputation strategy address ICE subject. column must named specified vars$strategy. Possible imputation strategies : \"MAR\": Missing Random. \"CIR\": Copy Increments Reference. \"CR\": Copy Reference. \"JR\": Jump Reference. \"LMCF\": Last Mean Carried Forward. explanations imputation strategies, see Carpenter, Roger, Kenward (2013), Cro et al (2021), Wolbers et al (2022). Please note user-defined imputation strategies can also set. data_ice argument necessary stage since (explained Wolbers et al (2022)), model fitted removing observations incompatible imputation model, .e. observed data data_ice[[vars$visit]] addressed imputation strategy different MAR excluded model fit. However observations discarded data imputation phase (performed function (impute()). summarize, stage pre-ICE data post-ICE data ICEs MAR imputation specified used. data_ice argument omitted, subject record within data_ice, assumed relevant subject's data pre-ICE missing visits imputed MAR assumption observed data used fit base imputation model. Please note ICE visit updated via update_strategy argument impute(); means subjects record data_ice always missing data imputed MAR assumption even strategy updated. vars argument named list specifies names key variables within data data_ice. list created set_vars() contains following named elements: subjid: name column data data_ice contains subject ids variable. visit: name column data data_ice contains visit variable. group: name column data contains group variable. outcome: name column data contains outcome variable. covariates: vector characters contains covariates included model (including interactions specified \"covariateName1*covariateName2\"``). covariates provided default model specification outcome ~ 1 + visit + groupwill used. Please note thegroup*visit` interaction included model default. strata: covariates used stratification variables bootstrap sampling. default vars$group set stratification variable. Needed method_condmean(type = \"bootstrap\") method_approxbayes(). strategy: name column data_ice contains subject-specific imputation strategy.","code":""},{"path":"/reference/draws.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Fit the base imputation model and get parameter estimates — draws","text":"James R Carpenter, James H Roger, Michael G Kenward. Analysis longitudinal trials protocol deviation: framework relevant, accessible assumptions, inference via multiple imputation. Journal Biopharmaceutical Statistics, 23(6):1352–1371, 2013. Suzie Cro, Tim P Morris, Michael G Kenward, James R Carpenter. Sensitivity analysis clinical trials missing continuous outcome data using controlled multiple imputation: practical guide. Statistics Medicine, 39(21):2815–2842, 2020. Roderick J. . Little Donald B. Rubin. Statistical Analysis Missing Data, Second Edition. John Wiley & Sons, Hoboken, New Jersey, 2002. [Section 10.2.3] Marcel Wolbers, Alessandro Noci, Paul Delmar, Craig Gower-Page, Sean Yiu, Jonathan W. Bartlett. Standard reference-based conditional mean imputation. https://arxiv.org/abs/2109.11162, 2022. Von Hippel, Paul T Bartlett, Jonathan W. Maximum likelihood multiple imputation: Faster imputations consistent standard errors without posterior draws. 2021.","code":""},{"path":[]},{"path":"/reference/encap_get_mmrm_sample.html","id":null,"dir":"Reference","previous_headings":"","what":"Encapsulate get_mmrm_sample — encap_get_mmrm_sample","title":"Encapsulate get_mmrm_sample — encap_get_mmrm_sample","text":"Function creates new wrapper function around get_mmrm_sample() arguments get_mmrm_sample() enclosed within new function. makes running parallel single process calls function smoother. particular function takes care exporting arguments required parallel process cluster","code":""},{"path":"/reference/encap_get_mmrm_sample.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Encapsulate get_mmrm_sample — encap_get_mmrm_sample","text":"","code":"encap_get_mmrm_sample(cl, longdata, method)"},{"path":"/reference/encap_get_mmrm_sample.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Encapsulate get_mmrm_sample — encap_get_mmrm_sample","text":"cl Either cluster get_cluster() NULL longdata longdata object longDataConstructor$new() method method object","code":""},{"path":[]},{"path":"/reference/eval_mmrm.html","id":null,"dir":"Reference","previous_headings":"","what":"Evaluate a call to mmrm — eval_mmrm","title":"Evaluate a call to mmrm — eval_mmrm","text":"utility function attempts evaluate call mmrm managing warnings errors thrown. particular function attempts catch warnings errors instead surfacing simply add additional element failed value TRUE. allows multiple calls made without program exiting.","code":""},{"path":"/reference/eval_mmrm.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Evaluate a call to mmrm — eval_mmrm","text":"","code":"eval_mmrm(expr)"},{"path":"/reference/eval_mmrm.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Evaluate a call to mmrm — eval_mmrm","text":"expr expression evaluated. call mmrm::mmrm().","code":""},{"path":"/reference/eval_mmrm.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Evaluate a call to mmrm — eval_mmrm","text":"function originally developed use glmmTMB needed hand-holding dropping false-positive warnings. important now kept around encase need catch false-positive warnings future.","code":""},{"path":[]},{"path":"/reference/eval_mmrm.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Evaluate a call to mmrm — eval_mmrm","text":"","code":"if (FALSE) { eval_mmrm({ mmrm::mmrm(formula, data) }) }"},{"path":"/reference/expand.html","id":null,"dir":"Reference","previous_headings":"","what":"Expand and fill in missing data.frame rows — expand","title":"Expand and fill in missing data.frame rows — expand","text":"functions essentially wrappers around base::expand.grid() ensure missing combinations data inserted data.frame imputation/fill methods updating covariate values newly created rows.","code":""},{"path":"/reference/expand.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Expand and fill in missing data.frame rows — expand","text":"","code":"expand(data, ...) fill_locf(data, vars, group = NULL, order = NULL) expand_locf(data, ..., vars, group, order)"},{"path":"/reference/expand.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Expand and fill in missing data.frame rows — expand","text":"data dataset expand fill . ... variables levels expanded (note duplicate entries levels result multiple rows level). vars character vector containing names variables need filled . group character vector containing names variables group performing LOCF imputation var. order character vector containing names additional variables sort data.frame performing LOCF.","code":""},{"path":"/reference/expand.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Expand and fill in missing data.frame rows — expand","text":"draws() function makes assumption subjects visits present data.frame covariate values non missing; expand(), fill_locf() expand_locf() utility functions support users ensuring data.frame's conform assumptions. expand() takes vectors expected levels data.frame expands combinations inserting missing rows data.frame. Note \"expanded\" variables cast factors. fill_locf() applies LOCF imputation named covariates fill NAs created insertion new rows expand() (though note distinction made existing NAs newly created NAs). Note data.frame sorted c(group, order) performing LOCF imputation; data.frame returned original sort order however. expand_locf() simple composition function fill_locf() expand() .e. fill_locf(expand(...)).","code":""},{"path":"/reference/expand.html","id":"missing-first-values","dir":"Reference","previous_headings":"","what":"Missing First Values","title":"Expand and fill in missing data.frame rows — expand","text":"fill_locf() function performs last observation carried forward imputation. natural consequence unable impute missing observations observation first value given subject / grouping. values deliberately imputed risks silent errors case time varying covariates. One solution first use expand_locf() just visit variable time varying covariates merge baseline covariates afterwards .e.","code":"library(dplyr) dat_expanded <- expand( data = dat, subject = c(\"pt1\", \"pt2\", \"pt3\", \"pt4\"), visit = c(\"vis1\", \"vis2\", \"vis3\") ) dat_filled <- dat_expanded %>% left_join(baseline_covariates, by = \"subject\")"},{"path":"/reference/expand.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Expand and fill in missing data.frame rows — expand","text":"","code":"if (FALSE) { dat_expanded <- expand( data = dat, subject = c(\"pt1\", \"pt2\", \"pt3\", \"pt4\"), visit = c(\"vis1\", \"vis2\", \"vis3\") ) dat_filled <- fill_loc( data = dat_expanded, vars = c(\"Sex\", \"Age\"), group = \"subject\", order = \"visit\" ) ## Or dat_filled <- expand_locf( data = dat, subject = c(\"pt1\", \"pt2\", \"pt3\", \"pt4\"), visit = c(\"vis1\", \"vis2\", \"vis3\"), vars = c(\"Sex\", \"Age\"), group = \"subject\", order = \"visit\" ) }"},{"path":"/reference/extract_covariates.html","id":null,"dir":"Reference","previous_headings":"","what":"Extract Variables from string vector — extract_covariates","title":"Extract Variables from string vector — extract_covariates","text":"Takes string including potentially model terms like * : extracts individual variables","code":""},{"path":"/reference/extract_covariates.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Extract Variables from string vector — extract_covariates","text":"","code":"extract_covariates(x)"},{"path":"/reference/extract_covariates.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Extract Variables from string vector — extract_covariates","text":"x string variable names potentially including interaction terms","code":""},{"path":"/reference/extract_covariates.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Extract Variables from string vector — extract_covariates","text":".e. c(\"v1\", \"v2\", \"v2*v3\", \"v1:v2\") becomes c(\"v1\", \"v2\", \"v3\")","code":""},{"path":"/reference/extract_data_nmar_as_na.html","id":null,"dir":"Reference","previous_headings":"","what":"Set to NA outcome values that would be MNAR if they were missing\n(i.e. which occur after an ICE handled using a reference-based imputation strategy) — extract_data_nmar_as_na","title":"Set to NA outcome values that would be MNAR if they were missing\n(i.e. which occur after an ICE handled using a reference-based imputation strategy) — extract_data_nmar_as_na","text":"Set NA outcome values MNAR missing (.e. occur ICE handled using reference-based imputation strategy)","code":""},{"path":"/reference/extract_data_nmar_as_na.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Set to NA outcome values that would be MNAR if they were missing\n(i.e. which occur after an ICE handled using a reference-based imputation strategy) — extract_data_nmar_as_na","text":"","code":"extract_data_nmar_as_na(longdata)"},{"path":"/reference/extract_data_nmar_as_na.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Set to NA outcome values that would be MNAR if they were missing\n(i.e. which occur after an ICE handled using a reference-based imputation strategy) — extract_data_nmar_as_na","text":"longdata R6 longdata object containing relevant input data information.","code":""},{"path":"/reference/extract_data_nmar_as_na.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Set to NA outcome values that would be MNAR if they were missing\n(i.e. which occur after an ICE handled using a reference-based imputation strategy) — extract_data_nmar_as_na","text":"data.frame containing longdata$get_data(longdata$ids), MNAR outcome values set NA.","code":""},{"path":"/reference/extract_draws.html","id":null,"dir":"Reference","previous_headings":"","what":"Extract draws from a stanfit object — extract_draws","title":"Extract draws from a stanfit object — extract_draws","text":"Extract draws stanfit object convert lists. function rstan::extract() returns draws given parameter array. function calls rstan::extract() extract draws stanfit object convert arrays lists.","code":""},{"path":"/reference/extract_draws.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Extract draws from a stanfit object — extract_draws","text":"","code":"extract_draws(stan_fit)"},{"path":"/reference/extract_draws.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Extract draws from a stanfit object — extract_draws","text":"stan_fit stanfit object.","code":""},{"path":"/reference/extract_draws.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Extract draws from a stanfit object — extract_draws","text":"named list length 2 containing: beta: list length equal number draws containing draws posterior distribution regression coefficients. sigma: list length equal number draws containing draws posterior distribution covariance matrices. element list list length equal 1 same_cov = TRUE equal number groups same_cov = FALSE.","code":""},{"path":"/reference/extract_imputed_df.html","id":null,"dir":"Reference","previous_headings":"","what":"Extract imputed dataset — extract_imputed_df","title":"Extract imputed dataset — extract_imputed_df","text":"Takes imputation object generated imputation_df() uses extract completed dataset longdata object created longDataConstructor(). Also applies delta transformation data.frame provided delta argument. See analyse() details structure data.frame. Subject IDs returned data.frame scrambled .e. original values.","code":""},{"path":"/reference/extract_imputed_df.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Extract imputed dataset — extract_imputed_df","text":"","code":"extract_imputed_df(imputation, ld, delta = NULL, idmap = FALSE)"},{"path":"/reference/extract_imputed_df.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Extract imputed dataset — extract_imputed_df","text":"imputation imputation object generated imputation_df(). ld longdata object generated longDataConstructor(). delta Either NULL data.frame. used offset outcome values imputed dataset. idmap Logical. TRUE attribute called \"idmap\" attached return object contains list maps old subject ids new subject ids.","code":""},{"path":"/reference/extract_imputed_df.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Extract imputed dataset — extract_imputed_df","text":"data.frame.","code":""},{"path":"/reference/extract_imputed_dfs.html","id":null,"dir":"Reference","previous_headings":"","what":"Extract imputed datasets — extract_imputed_dfs","title":"Extract imputed datasets — extract_imputed_dfs","text":"Extracts imputed datasets contained within imputations object generated impute().","code":""},{"path":"/reference/extract_imputed_dfs.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Extract imputed datasets — extract_imputed_dfs","text":"","code":"extract_imputed_dfs( imputations, index = seq_along(imputations$imputations), delta = NULL, idmap = FALSE )"},{"path":"/reference/extract_imputed_dfs.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Extract imputed datasets — extract_imputed_dfs","text":"imputations imputations object created impute(). index indexes imputed datasets return. default, datasets within imputations object returned. delta data.frame containing delta transformation applied imputed dataset. See analyse() details format specification data.frame. idmap Logical. subject IDs imputed data.frame's replaced new IDs ensure unique. Setting argument TRUE attaches attribute, called idmap, returned data.frame's provide map new subject IDs old subject IDs.","code":""},{"path":"/reference/extract_imputed_dfs.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Extract imputed datasets — extract_imputed_dfs","text":"list data.frames equal length index argument.","code":""},{"path":[]},{"path":"/reference/extract_imputed_dfs.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Extract imputed datasets — extract_imputed_dfs","text":"","code":"if (FALSE) { extract_imputed_dfs(imputeObj) extract_imputed_dfs(imputeObj, c(1:3)) }"},{"path":"/reference/extract_params.html","id":null,"dir":"Reference","previous_headings":"","what":"Extract parameters from a MMRM model — extract_params","title":"Extract parameters from a MMRM model — extract_params","text":"Extracts beta sigma coefficients MMRM model created mmrm::mmrm().","code":""},{"path":"/reference/extract_params.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Extract parameters from a MMRM model — extract_params","text":"","code":"extract_params(fit)"},{"path":"/reference/extract_params.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Extract parameters from a MMRM model — extract_params","text":"fit object created mmrm::mmrm()","code":""},{"path":"/reference/fit_mcmc.html","id":null,"dir":"Reference","previous_headings":"","what":"Fit the base imputation model using a Bayesian approach — fit_mcmc","title":"Fit the base imputation model using a Bayesian approach — fit_mcmc","text":"fit_mcmc() fits base imputation model using Bayesian approach. done MCMC method implemented stan run using function rstan::sampling(). function returns draws posterior distribution model parameters stanfit object. Additionally performs multiple diagnostics checks chain returns warnings case detected issues.","code":""},{"path":"/reference/fit_mcmc.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Fit the base imputation model using a Bayesian approach — fit_mcmc","text":"","code":"fit_mcmc(designmat, outcome, group, subjid, visit, method, quiet = FALSE)"},{"path":"/reference/fit_mcmc.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Fit the base imputation model using a Bayesian approach — fit_mcmc","text":"designmat design matrix fixed effects. outcome response variable. Must numeric. group Character vector containing group variable. subjid Character vector containing subjects IDs. visit Character vector containing visit variable. method method object generated method_bayes(). quiet Specify whether stan sampling log printed console.","code":""},{"path":"/reference/fit_mcmc.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Fit the base imputation model using a Bayesian approach — fit_mcmc","text":"named list composed following: samples: named list containing draws parameter. corresponds output extract_draws(). fit: stanfit object.","code":""},{"path":"/reference/fit_mcmc.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Fit the base imputation model using a Bayesian approach — fit_mcmc","text":"Bayesian model assumes multivariate normal likelihood function weakly-informative priors model parameters: particular, uniform priors assumed regression coefficients inverse-Wishart priors covariance matrices. chain initialized using REML parameter estimates MMRM starting values. function performs following steps: Fit MMRM using REML approach. Prepare input data MCMC fit described data{} block Stan file. See prepare_stan_data() details. Run MCMC according input arguments using starting values REML parameter estimates estimated point 1. Performs diagnostics checks MCMC. See check_mcmc() details. Extract draws model fit. chains perform method$n_samples draws keeping one every method$burn_between iterations. Additionally first method$burn_in iterations discarded. total number iterations method$burn_in + method$burn_between*method$n_samples. purpose method$burn_in ensure samples drawn stationary distribution Markov Chain. method$burn_between aims keep draws uncorrelated .","code":""},{"path":"/reference/fit_mmrm.html","id":null,"dir":"Reference","previous_headings":"","what":"Fit a MMRM model — fit_mmrm","title":"Fit a MMRM model — fit_mmrm","text":"Fits MMRM model allowing different covariance structures using mmrm::mmrm(). Returns list key model parameters beta, sigma additional element failed indicating whether fit failed converge. fit fail converge beta sigma present.","code":""},{"path":"/reference/fit_mmrm.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Fit a MMRM model — fit_mmrm","text":"","code":"fit_mmrm( designmat, outcome, subjid, visit, group, cov_struct = c(\"us\", \"toep\", \"cs\", \"ar1\"), REML = TRUE, same_cov = TRUE )"},{"path":"/reference/fit_mmrm.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Fit a MMRM model — fit_mmrm","text":"designmat data.frame matrix containing covariates use MMRM model. Dummy variables must already expanded , .e. via stats::model.matrix(). contain missing values outcome numeric vector. outcome value regressed MMRM model. subjid character / factor vector. subject identifier used link separate visits belong subject. visit character / factor vector. Indicates visit outcome value occurred . group character / factor vector. Indicates treatment group patient belongs . cov_struct character value. Specifies covariance structure use. Must one \"us\", \"toep\", \"cs\" \"ar1\" REML logical. Specifies whether restricted maximum likelihood used same_cov logical. Used specify shared individual covariance matrix used per group","code":""},{"path":"/reference/generate_data_single.html","id":null,"dir":"Reference","previous_headings":"","what":"Generate data for a single group — generate_data_single","title":"Generate data for a single group — generate_data_single","text":"Generate data single group","code":""},{"path":"/reference/generate_data_single.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Generate data for a single group — generate_data_single","text":"","code":"generate_data_single(pars_group, strategy_fun = NULL, distr_pars_ref = NULL)"},{"path":"/reference/generate_data_single.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Generate data for a single group — generate_data_single","text":"pars_group simul_pars object generated set_simul_pars(). specifies simulation parameters given group. strategy_fun Function implementing trajectories intercurrent event (ICE). Must one getStrategies(). See getStrategies() details. NULL post-ICE outcomes untouched. distr_pars_ref Optional. Named list containing simulation parameters reference arm. contains following elements: mu: Numeric vector indicating mean outcome trajectory assuming ICEs. include outcome baseline. sigma Covariance matrix outcome trajectory assuming ICEs. NULL, parameters inherited pars_group.","code":""},{"path":"/reference/generate_data_single.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Generate data for a single group — generate_data_single","text":"data.frame containing simulated data. includes following variables: id: Factor variable specifies id subject. visit: Factor variable specifies visit assessment. Visit 0 denotes baseline visit. group: Factor variable specifies treatment group subject belongs . outcome_bl: Numeric variable specifies baseline outcome. outcome_noICE: Numeric variable specifies longitudinal outcome assuming ICEs. ind_ice1: Binary variable takes value 1 corresponding visit affected ICE1 0 otherwise. dropout_ice1: Binary variable takes value 1 corresponding visit affected drop-following ICE1 0 otherwise. ind_ice2: Binary variable takes value 1 corresponding visit affected ICE2. outcome: Numeric variable specifies longitudinal outcome including ICE1, ICE2 intermittent missing values.","code":""},{"path":[]},{"path":"/reference/getStrategies.html","id":null,"dir":"Reference","previous_headings":"","what":"Get imputation strategies — getStrategies","title":"Get imputation strategies — getStrategies","text":"Returns list defining imputation strategies used create multivariate normal distribution parameters merging source group reference group per patient.","code":""},{"path":"/reference/getStrategies.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Get imputation strategies — getStrategies","text":"","code":"getStrategies(...)"},{"path":"/reference/getStrategies.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Get imputation strategies — getStrategies","text":"... User defined methods added return list. Input must function.","code":""},{"path":"/reference/getStrategies.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Get imputation strategies — getStrategies","text":"default Jump Reference (JR), Copy Reference (CR), Copy Increments Reference (CIR), Last Mean Carried Forward (LMCF) Missing Random (MAR) defined. user can define strategy functions (overwrite pre-defined ones) specifying named input function .e. NEW = function(...) .... exception MAR overwritten. user defined functions must take 3 inputs: pars_group, pars_ref index_mar. pars_group pars_ref lists elements mu sigma representing multivariate normal distribution parameters subject's current group reference group respectively. index_mar logical vector specifying visits subject met MAR assumption . function must return list elements mu sigma. See implementation strategy_JR() example.","code":""},{"path":"/reference/getStrategies.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Get imputation strategies — getStrategies","text":"","code":"if (FALSE) { getStrategies() getStrategies( NEW = function(pars_group, pars_ref, index_mar) code , JR = function(pars_group, pars_ref, index_mar) more_code ) }"},{"path":"/reference/get_ESS.html","id":null,"dir":"Reference","previous_headings":"","what":"Extract the Effective Sample Size (ESS) from a stanfit object — get_ESS","title":"Extract the Effective Sample Size (ESS) from a stanfit object — get_ESS","text":"Extract Effective Sample Size (ESS) stanfit object","code":""},{"path":"/reference/get_ESS.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Extract the Effective Sample Size (ESS) from a stanfit object — get_ESS","text":"","code":"get_ESS(stan_fit)"},{"path":"/reference/get_ESS.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Extract the Effective Sample Size (ESS) from a stanfit object — get_ESS","text":"stan_fit stanfit object.","code":""},{"path":"/reference/get_ESS.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Extract the Effective Sample Size (ESS) from a stanfit object — get_ESS","text":"named vector containing ESS parameter model.","code":""},{"path":"/reference/get_bootstrap_stack.html","id":null,"dir":"Reference","previous_headings":"","what":"Creates a stack object populated with bootstrapped samples — get_bootstrap_stack","title":"Creates a stack object populated with bootstrapped samples — get_bootstrap_stack","text":"Function creates Stack() object populated stack bootstrap samples based upon method$n_samples","code":""},{"path":"/reference/get_bootstrap_stack.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Creates a stack object populated with bootstrapped samples — get_bootstrap_stack","text":"","code":"get_bootstrap_stack(longdata, method, stack = Stack$new())"},{"path":"/reference/get_bootstrap_stack.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Creates a stack object populated with bootstrapped samples — get_bootstrap_stack","text":"longdata longDataConstructor() object method method object stack Stack() object (exposed unit testing purposes)","code":""},{"path":"/reference/get_cluster.html","id":null,"dir":"Reference","previous_headings":"","what":"Create cluster — get_cluster","title":"Create cluster — get_cluster","text":"Create cluster","code":""},{"path":"/reference/get_cluster.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create cluster — get_cluster","text":"","code":"get_cluster(ncores = 1)"},{"path":"/reference/get_cluster.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create cluster — get_cluster","text":"ncores Number parallel processes use ncores 1 function return NULL function spawns PSOCK cluster. Ensures rbmi assert_that loaded sub-processes","code":""},{"path":"/reference/get_conditional_parameters.html","id":null,"dir":"Reference","previous_headings":"","what":"Derive conditional multivariate normal parameters — get_conditional_parameters","title":"Derive conditional multivariate normal parameters — get_conditional_parameters","text":"Takes parameters multivariate normal distribution observed values calculate conditional distribution unobserved values.","code":""},{"path":"/reference/get_conditional_parameters.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Derive conditional multivariate normal parameters — get_conditional_parameters","text":"","code":"get_conditional_parameters(pars, values)"},{"path":"/reference/get_conditional_parameters.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Derive conditional multivariate normal parameters — get_conditional_parameters","text":"pars list elements mu sigma defining mean vector covariance matrix respectively. values vector observed values condition , must length pars$mu. Missing values must represented NA.","code":""},{"path":"/reference/get_conditional_parameters.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Derive conditional multivariate normal parameters — get_conditional_parameters","text":"list conditional distribution parameters: mu - conditional mean vector. sigma - conditional covariance matrix.","code":""},{"path":"/reference/get_delta_template.html","id":null,"dir":"Reference","previous_headings":"","what":"Get delta utility variables — get_delta_template","title":"Get delta utility variables — get_delta_template","text":"function creates default delta template (1 row per subject per visit) extracts utility information users need define logic defining delta. See delta_template() full details.","code":""},{"path":"/reference/get_delta_template.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Get delta utility variables — get_delta_template","text":"","code":"get_delta_template(imputations)"},{"path":"/reference/get_delta_template.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Get delta utility variables — get_delta_template","text":"imputations imputations object created impute().","code":""},{"path":"/reference/get_draws_mle.html","id":null,"dir":"Reference","previous_headings":"","what":"Fit the base imputation model on bootstrap samples — get_draws_mle","title":"Fit the base imputation model on bootstrap samples — get_draws_mle","text":"Fit base imputation model using ML/REML approach given number bootstrap samples specified method$n_samples. Returns parameter estimates model fit.","code":""},{"path":"/reference/get_draws_mle.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Fit the base imputation model on bootstrap samples — get_draws_mle","text":"","code":"get_draws_mle( longdata, method, sample_stack, n_target_samples, first_sample_orig, use_samp_ids, failure_limit = 0, ncores = 1, quiet = FALSE )"},{"path":"/reference/get_draws_mle.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Fit the base imputation model on bootstrap samples — get_draws_mle","text":"longdata R6 longdata object containing relevant input data information. method method object generated either method_approxbayes() method_condmean() argument type = \"bootstrap\". sample_stack stack object containing subject ids used mmrm iteration. n_target_samples Number samples needed created first_sample_orig Logical. TRUE function returns method$n_samples + 1 samples first sample contains parameter estimates original dataset method$n_samples samples contain parameter estimates bootstrap samples. FALSE function returns method$n_samples samples containing parameter estimates bootstrap samples. use_samp_ids Logical. TRUE, sampled subject ids returned. Otherwise subject ids original dataset returned. values used tell impute() subjects used derive imputed dataset. failure_limit Number failed samples allowed throwing error ncores Number processes parallelise job quiet Logical, TRUE suppress printing progress information printed console.","code":""},{"path":"/reference/get_draws_mle.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Fit the base imputation model on bootstrap samples — get_draws_mle","text":"draws object named list containing following: data: R6 longdata object containing relevant input data information. method: method object generated either method_bayes(), method_approxbayes() method_condmean(). samples: list containing estimated parameters interest. element samples named list containing following: ids: vector characters containing ids subjects included original dataset. beta: numeric vector estimated regression coefficients. sigma: list estimated covariance matrices (one level vars$group). theta: numeric vector transformed covariances. failed: Logical. TRUE model fit failed. ids_samp: vector characters containing ids subjects included given sample. fit: method_bayes() chosen, returns MCMC Stan fit object. Otherwise NULL. n_failures: absolute number failures model fit. Relevant method_condmean(type = \"bootstrap\"), method_approxbayes() method_bmlmi(). formula: fixed effects formula object used model specification.","code":""},{"path":"/reference/get_draws_mle.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Fit the base imputation model on bootstrap samples — get_draws_mle","text":"function takes Stack object contains multiple lists patient ids. function takes Stack pulls set ids constructs dataset just consisting patients (.e. potentially bootstrap jackknife sample). function fits MMRM model dataset create sample object. function repeats process n_target_samples reached. failure_limit samples fail converge function throws error. reaching desired number samples function generates returns draws object.","code":""},{"path":"/reference/get_ests_bmlmi.html","id":null,"dir":"Reference","previous_headings":"","what":"Von Hippel and Bartlett pooling of BMLMI method — get_ests_bmlmi","title":"Von Hippel and Bartlett pooling of BMLMI method — get_ests_bmlmi","text":"Compute pooled point estimates, standard error degrees freedom according Von Hippel Bartlett formula Bootstrapped Maximum Likelihood Multiple Imputation (BMLMI).","code":""},{"path":"/reference/get_ests_bmlmi.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Von Hippel and Bartlett pooling of BMLMI method — get_ests_bmlmi","text":"","code":"get_ests_bmlmi(ests, D)"},{"path":"/reference/get_ests_bmlmi.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Von Hippel and Bartlett pooling of BMLMI method — get_ests_bmlmi","text":"ests numeric vector containing estimates analysis imputed datasets. D numeric representing number imputations bootstrap sample BMLMI method.","code":""},{"path":"/reference/get_ests_bmlmi.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Von Hippel and Bartlett pooling of BMLMI method — get_ests_bmlmi","text":"list containing point estimate, standard error degrees freedom.","code":""},{"path":"/reference/get_ests_bmlmi.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Von Hippel and Bartlett pooling of BMLMI method — get_ests_bmlmi","text":"ests must provided following order: firsts D elements related analyses random imputation one bootstrap sample. second set D elements (.e. D+1 2*D) related second bootstrap sample .","code":""},{"path":"/reference/get_ests_bmlmi.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Von Hippel and Bartlett pooling of BMLMI method — get_ests_bmlmi","text":"Von Hippel, Paul T Bartlett, Jonathan W8. Maximum likelihood multiple imputation: Faster imputations consistent standard errors without posterior draws. 2021","code":""},{"path":"/reference/get_example_data.html","id":null,"dir":"Reference","previous_headings":"","what":"Simulate a realistic example dataset — get_example_data","title":"Simulate a realistic example dataset — get_example_data","text":"Simulate realistic example dataset using simulate_data() hard-coded values input arguments.","code":""},{"path":"/reference/get_example_data.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Simulate a realistic example dataset — get_example_data","text":"","code":"get_example_data()"},{"path":"/reference/get_example_data.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Simulate a realistic example dataset — get_example_data","text":"get_example_data() simulates 1:1 randomized trial active drug (intervention) versus placebo (control) 100 subjects per group 6 post-baseline assessments (bi-monthly visits 12 months). One intercurrent event corresponding treatment discontinuation also simulated. Specifically, data simulated following assumptions: mean outcome trajectory placebo group increases linearly 50 baseline (visit 0) 60 visit 6, .e. slope 10 points/year. mean outcome trajectory intervention group identical placebo group visit 2. visit 2 onward, slope decreases 50% 5 points/year. covariance structure baseline follow-values groups implied random intercept slope model standard deviation 5 intercept slope, correlation 0.25. addition, independent residual error standard deviation 2.5 added assessment. probability study drug discontinuation visit calculated according logistic model depends observed outcome visit. Specifically, visit-wise discontinuation probability 2% 3% control intervention group, respectively, specified case observed outcome equal 50 (mean value baseline). odds discontinuation simulated increase +10% +1 point increase observed outcome. Study drug discontinuation simulated effect mean trajectory placebo group. intervention group, subjects discontinue follow slope mean trajectory placebo group time point onward. compatible copy increments reference (CIR) assumption. Study drop-study drug discontinuation visit occurs probability 50% leading missing outcome data time point onward.","code":""},{"path":[]},{"path":"/reference/get_jackknife_stack.html","id":null,"dir":"Reference","previous_headings":"","what":"Creates a stack object populated with jackknife samples — get_jackknife_stack","title":"Creates a stack object populated with jackknife samples — get_jackknife_stack","text":"Function creates Stack() object populated stack jackknife samples based upon","code":""},{"path":"/reference/get_jackknife_stack.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Creates a stack object populated with jackknife samples — get_jackknife_stack","text":"","code":"get_jackknife_stack(longdata, method, stack = Stack$new())"},{"path":"/reference/get_jackknife_stack.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Creates a stack object populated with jackknife samples — get_jackknife_stack","text":"longdata longDataConstructor() object method method object stack Stack() object (exposed unit testing purposes)","code":""},{"path":"/reference/get_mmrm_sample.html","id":null,"dir":"Reference","previous_headings":"","what":"Fit MMRM and returns parameter estimates — get_mmrm_sample","title":"Fit MMRM and returns parameter estimates — get_mmrm_sample","text":"get_mmrm_sample fits base imputation model using ML/REML approach. Returns parameter estimates fit.","code":""},{"path":"/reference/get_mmrm_sample.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Fit MMRM and returns parameter estimates — get_mmrm_sample","text":"","code":"get_mmrm_sample(ids, longdata, method)"},{"path":"/reference/get_mmrm_sample.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Fit MMRM and returns parameter estimates — get_mmrm_sample","text":"ids vector characters containing ids subjects. longdata R6 longdata object containing relevant input data information. method method object generated either method_approxbayes() method_condmean().","code":""},{"path":"/reference/get_mmrm_sample.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Fit MMRM and returns parameter estimates — get_mmrm_sample","text":"named list class sample_single. contains following: ids vector characters containing ids subjects included original dataset. beta numeric vector estimated regression coefficients. sigma list estimated covariance matrices (one level vars$group). theta numeric vector transformed covariances. failed logical. TRUE model fit failed. ids_samp vector characters containing ids subjects included given sample.","code":""},{"path":"/reference/get_pattern_groups.html","id":null,"dir":"Reference","previous_headings":"","what":"Determine patients missingness group — get_pattern_groups","title":"Determine patients missingness group — get_pattern_groups","text":"Takes design matrix multiple rows per subject returns dataset 1 row per subject new column pgroup indicating group patient belongs (based upon missingness pattern treatment group)","code":""},{"path":"/reference/get_pattern_groups.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Determine patients missingness group — get_pattern_groups","text":"","code":"get_pattern_groups(ddat)"},{"path":"/reference/get_pattern_groups.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Determine patients missingness group — get_pattern_groups","text":"ddat data.frame columns subjid, visit, group, is_avail","code":""},{"path":"/reference/get_pattern_groups.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Determine patients missingness group — get_pattern_groups","text":"column is_avail must character numeric 0 1","code":""},{"path":"/reference/get_pattern_groups_unique.html","id":null,"dir":"Reference","previous_headings":"","what":"Get Pattern Summary — get_pattern_groups_unique","title":"Get Pattern Summary — get_pattern_groups_unique","text":"Takes dataset pattern information creates summary dataset just 1 row per pattern","code":""},{"path":"/reference/get_pattern_groups_unique.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Get Pattern Summary — get_pattern_groups_unique","text":"","code":"get_pattern_groups_unique(patterns)"},{"path":"/reference/get_pattern_groups_unique.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Get Pattern Summary — get_pattern_groups_unique","text":"patterns data.frame columns pgroup, pattern group","code":""},{"path":"/reference/get_pattern_groups_unique.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Get Pattern Summary — get_pattern_groups_unique","text":"column pgroup must numeric vector indicating pattern group patient belongs column pattern must character string 0's 1's. must identical rows within pgroup column group must character / numeric vector indicating covariance group observation belongs . must identical within pgroup","code":""},{"path":"/reference/get_pool_components.html","id":null,"dir":"Reference","previous_headings":"","what":"Expected Pool Components — get_pool_components","title":"Expected Pool Components — get_pool_components","text":"Returns elements expected contained analyse object depending analysis method specified.","code":""},{"path":"/reference/get_pool_components.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Expected Pool Components — get_pool_components","text":"","code":"get_pool_components(x)"},{"path":"/reference/get_pool_components.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Expected Pool Components — get_pool_components","text":"x Character name analysis method, must one either \"rubin\", \"jackknife\", \"bootstrap\" \"bmlmi\".","code":""},{"path":"/reference/get_visit_distribution_parameters.html","id":null,"dir":"Reference","previous_headings":"","what":"Derive visit distribution parameters — get_visit_distribution_parameters","title":"Derive visit distribution parameters — get_visit_distribution_parameters","text":"Takes patient level data beta coefficients expands get patient specific estimate visit distribution parameters mu sigma. Returns values specific format expected downstream functions imputation process (namely list(list(mu = ..., sigma = ...), list(mu = ..., sigma = ...))).","code":""},{"path":"/reference/get_visit_distribution_parameters.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Derive visit distribution parameters — get_visit_distribution_parameters","text":"","code":"get_visit_distribution_parameters(dat, beta, sigma)"},{"path":"/reference/get_visit_distribution_parameters.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Derive visit distribution parameters — get_visit_distribution_parameters","text":"dat Patient level dataset, must 1 row per visit. Column order must order beta. number columns must match length beta beta List model beta coefficients. 1 element sample e.g. 3 samples models 4 beta coefficients argument form list( c(1,2,3,4) , c(5,6,7,8), c(9,10,11,12)). elements beta must length must length order dat. sigma List sigma. Must number entries beta.","code":""},{"path":"/reference/has_class.html","id":null,"dir":"Reference","previous_headings":"","what":"Does object have a class ? — has_class","title":"Does object have a class ? — has_class","text":"Utility function see object particular class. Useful know many classes object may .","code":""},{"path":"/reference/has_class.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Does object have a class ? — has_class","text":"","code":"has_class(x, cls)"},{"path":"/reference/has_class.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Does object have a class ? — has_class","text":"x object want check class . cls class want know .","code":""},{"path":"/reference/has_class.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Does object have a class ? — has_class","text":"TRUE object class. FALSE object class.","code":""},{"path":"/reference/ife.html","id":null,"dir":"Reference","previous_headings":"","what":"if else — ife","title":"if else — ife","text":"wrapper around () else() prevent unexpected interactions ifelse() factor variables","code":""},{"path":"/reference/ife.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"if else — ife","text":"","code":"ife(x, a, b)"},{"path":"/reference/ife.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"if else — ife","text":"x True / False value return True b value return False","code":""},{"path":"/reference/ife.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"if else — ife","text":"default ifelse() convert factor variables numeric values often undesirable. connivance function avoids problem","code":""},{"path":"/reference/imputation_df.html","id":null,"dir":"Reference","previous_headings":"","what":"Create a valid imputation_df object — imputation_df","title":"Create a valid imputation_df object — imputation_df","text":"Create valid imputation_df object","code":""},{"path":"/reference/imputation_df.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create a valid imputation_df object — imputation_df","text":"","code":"imputation_df(...)"},{"path":"/reference/imputation_df.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create a valid imputation_df object — imputation_df","text":"... list imputation_single.","code":""},{"path":"/reference/imputation_list_df.html","id":null,"dir":"Reference","previous_headings":"","what":"List of imputations_df — imputation_list_df","title":"List of imputations_df — imputation_list_df","text":"container multiple imputation_df's","code":""},{"path":"/reference/imputation_list_df.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"List of imputations_df — imputation_list_df","text":"","code":"imputation_list_df(...)"},{"path":"/reference/imputation_list_df.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"List of imputations_df — imputation_list_df","text":"... objects class imputation_df","code":""},{"path":"/reference/imputation_list_single.html","id":null,"dir":"Reference","previous_headings":"","what":"A collection of imputation_singles() grouped by a single subjid ID — imputation_list_single","title":"A collection of imputation_singles() grouped by a single subjid ID — imputation_list_single","text":"collection imputation_singles() grouped single subjid ID","code":""},{"path":"/reference/imputation_list_single.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"A collection of imputation_singles() grouped by a single subjid ID — imputation_list_single","text":"","code":"imputation_list_single(imputations, D = 1)"},{"path":"/reference/imputation_list_single.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"A collection of imputation_singles() grouped by a single subjid ID — imputation_list_single","text":"imputations list imputation_single() objects ordered repetitions grouped sequentially D number repetitions performed determines many columns imputation matrix constructor function create imputation_list_single object contains matrix imputation_single() objects grouped single id. matrix split D columns (.e. non-bmlmi methods always 1) id attribute determined extracting id attribute contributing imputation_single() objects. error throw multiple id detected","code":""},{"path":"/reference/imputation_single.html","id":null,"dir":"Reference","previous_headings":"","what":"Create a valid imputation_single object — imputation_single","title":"Create a valid imputation_single object — imputation_single","text":"Create valid imputation_single object","code":""},{"path":"/reference/imputation_single.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create a valid imputation_single object — imputation_single","text":"","code":"imputation_single(id, values)"},{"path":"/reference/imputation_single.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create a valid imputation_single object — imputation_single","text":"id character string specifying subject id. values numeric vector indicating imputed values.","code":""},{"path":"/reference/impute.html","id":null,"dir":"Reference","previous_headings":"","what":"Create imputed datasets — impute","title":"Create imputed datasets — impute","text":"impute() creates imputed datasets based upon data options specified call draws(). One imputed dataset created per \"sample\" created draws().","code":""},{"path":"/reference/impute.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create imputed datasets — impute","text":"","code":"impute( draws, references = NULL, update_strategy = NULL, strategies = getStrategies() ) # S3 method for random impute( draws, references = NULL, update_strategy = NULL, strategies = getStrategies() ) # S3 method for condmean impute( draws, references = NULL, update_strategy = NULL, strategies = getStrategies() )"},{"path":"/reference/impute.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create imputed datasets — impute","text":"draws draws object created draws(). references named vector. Identifies references used reference-based imputation methods. form c(\"Group1\" = \"Reference1\", \"Group2\" = \"Reference2\"). NULL (default), references assumed form c(\"Group1\" = \"Group1\", \"Group2\" = \"Group2\"). argument NULL imputation strategy (defined data_ice[[vars$strategy]] call draws) MAR set. update_strategy optional data.frame. Updates imputation method originally set via data_ice option draws(). See details section information. strategies named list functions. Defines imputation functions used. names list mirror values specified strategy column data_ice. Default = getStrategies(). See getStrategies() details.","code":""},{"path":"/reference/impute.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Create imputed datasets — impute","text":"impute() uses imputation model parameter estimates, generated draws(), first calculate marginal (multivariate normal) distribution subject's longitudinal outcome variable depending covariate values. subjects intercurrent events (ICEs) handled using non-MAR methods, marginal distribution updated depending time first visit affected ICE, chosen imputation strategy chosen reference group described Carpenter, Roger, Kenward (2013) . subject's imputation distribution used imputing missing values defined marginal distribution conditional observed outcome values. One dataset generated per set parameter estimates provided draws(). exact manner missing values imputed conditional imputation distribution depends method object provided draws(), particular: Bayes & Approximate Bayes: imputed dataset contains 1 row per subject & visit original dataset missing values imputed taking single random sample conditional imputation distribution. Conditional Mean: imputed dataset contains 1 row per subject & visit bootstrapped jackknife dataset used generate corresponding parameter estimates draws(). Missing values imputed using mean conditional imputation distribution. Please note first imputed dataset refers conditional mean imputation original dataset whereas subsequent imputed datasets refer conditional mean imputations bootstrap jackknife samples, respectively, original data. Bootstrapped Maximum Likelihood MI (BMLMI): performs D random imputations bootstrapped dataset used generate corresponding parameter estimates draws(). total number B*D imputed datasets provided, B number bootstrapped datasets. Missing values imputed taking random sample conditional imputation distribution. update_strategy argument can used update imputation strategy originally set via data_ice option draws(). avoids re-run draws() function changing imputation strategy certain circumstances (detailed ). data.frame provided update_strategy argument must contain two columns, one subject ID another imputation strategy, whose names defined vars argument specified call draws(). Please note argument allows update imputation strategy arguments time first visit affected ICE. key limitation functionality one can switch MAR non-MAR strategy (vice versa) subjects without observed post-ICE data. reason change affect whether post-ICE data included base imputation model (explained help draws()). example, subject ICE \"Visit 2\" observed/known values \"Visit 3\" function throw error one tries switch strategy MAR non-MAR strategy. contrast, switching non-MAR MAR strategy, whilst valid, raise warning usable data utilised imputation model.","code":""},{"path":"/reference/impute.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Create imputed datasets — impute","text":"James R Carpenter, James H Roger, Michael G Kenward. Analysis longitudinal trials protocol deviation: framework relevant, accessible assumptions, inference via multiple imputation. Journal Biopharmaceutical Statistics, 23(6):1352–1371, 2013. [Section 4.2 4.3]","code":""},{"path":"/reference/impute.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Create imputed datasets — impute","text":"","code":"if (FALSE) { impute( draws = drawobj, references = c(\"Trt\" = \"Placebo\", \"Placebo\" = \"Placebo\") ) new_strategy <- data.frame( subjid = c(\"Pt1\", \"Pt2\"), strategy = c(\"MAR\", \"JR\") ) impute( draws = drawobj, references = c(\"Trt\" = \"Placebo\", \"Placebo\" = \"Placebo\"), update_strategy = new_strategy ) }"},{"path":"/reference/impute_data_individual.html","id":null,"dir":"Reference","previous_headings":"","what":"Impute data for a single subject — impute_data_individual","title":"Impute data for a single subject — impute_data_individual","text":"function performs imputation single subject time implementing process detailed impute().","code":""},{"path":"/reference/impute_data_individual.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Impute data for a single subject — impute_data_individual","text":"","code":"impute_data_individual( id, index, beta, sigma, data, references, strategies, condmean, n_imputations = 1 )"},{"path":"/reference/impute_data_individual.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Impute data for a single subject — impute_data_individual","text":"id Character string identifying subject. index sample indexes subject belongs e.g c(1,1,1,2,2,4). beta list beta coefficients sample, .e. beta[[1]] set beta coefficients first sample. sigma list sigma coefficients sample split group .e. sigma[[1]][[\"\"]] give sigma coefficients group first sample. data longdata object created longDataConstructor() references named vector. Identifies references used generating imputed values. form c(\"Group\" = \"Reference\", \"Group\" = \"Reference\"). strategies named list functions. Defines imputation functions used. names list mirror values specified method column data_ice. Default = getStrategies(). See getStrategies() details. condmean Logical. TRUE impute using conditional mean values, FALSE impute taking random draw multivariate normal distribution. n_imputations condmean = FALSE numeric representing number random imputations performed sample. Default 1 (one random imputation per sample).","code":""},{"path":"/reference/impute_data_individual.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Impute data for a single subject — impute_data_individual","text":"Note function performs required imputations subject time. .e. subject included samples 1,3,5,9 imputations (using sample-dependent imputation model parameters) performed one step order avoid look subjects's covariates expanding design matrix multiple times (computationally expensive). function also supports subject belonging sample multiple times, .e. 1,1,2,3,5,5, typically occur bootstrapped datasets.","code":""},{"path":"/reference/impute_internal.html","id":null,"dir":"Reference","previous_headings":"","what":"Create imputed datasets — impute_internal","title":"Create imputed datasets — impute_internal","text":"work horse function implements functionality impute. See user level function impute() details.","code":""},{"path":"/reference/impute_internal.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create imputed datasets — impute_internal","text":"","code":"impute_internal( draws, references = NULL, update_strategy, strategies, condmean )"},{"path":"/reference/impute_internal.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create imputed datasets — impute_internal","text":"draws draws object created draws(). references named vector. Identifies references used reference-based imputation methods. form c(\"Group1\" = \"Reference1\", \"Group2\" = \"Reference2\"). NULL (default), references assumed form c(\"Group1\" = \"Group1\", \"Group2\" = \"Group2\"). argument NULL imputation strategy (defined data_ice[[vars$strategy]] call draws) MAR set. update_strategy optional data.frame. Updates imputation method originally set via data_ice option draws(). See details section information. strategies named list functions. Defines imputation functions used. names list mirror values specified strategy column data_ice. Default = getStrategies(). See getStrategies() details. condmean logical. TRUE impute using conditional mean values, values impute taking random draw multivariate normal distribution.","code":""},{"path":"/reference/impute_outcome.html","id":null,"dir":"Reference","previous_headings":"","what":"Sample outcome value — impute_outcome","title":"Sample outcome value — impute_outcome","text":"Draws random sample multivariate normal distribution.","code":""},{"path":"/reference/impute_outcome.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Sample outcome value — impute_outcome","text":"","code":"impute_outcome(conditional_parameters, n_imputations = 1, condmean = FALSE)"},{"path":"/reference/impute_outcome.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Sample outcome value — impute_outcome","text":"conditional_parameters list elements mu sigma contain mean vector covariance matrix sample . n_imputations numeric representing number random samples multivariate normal distribution performed. Default 1. condmean conditional mean imputation performed (opposed random sampling)","code":""},{"path":"/reference/invert.html","id":null,"dir":"Reference","previous_headings":"","what":"invert — invert","title":"invert — invert","text":"Utility function used replicated purrr::transpose. Turns list inside .","code":""},{"path":"/reference/invert.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"invert — invert","text":"","code":"invert(x)"},{"path":"/reference/invert.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"invert — invert","text":"x list","code":""},{"path":"/reference/invert_indexes.html","id":null,"dir":"Reference","previous_headings":"","what":"Invert and derive indexes — invert_indexes","title":"Invert and derive indexes — invert_indexes","text":"Takes list elements creates new list containing 1 entry per unique element value containing indexes original elements occurred .","code":""},{"path":"/reference/invert_indexes.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Invert and derive indexes — invert_indexes","text":"","code":"invert_indexes(x)"},{"path":"/reference/invert_indexes.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Invert and derive indexes — invert_indexes","text":"x list elements invert calculate index (see details).","code":""},{"path":"/reference/invert_indexes.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Invert and derive indexes — invert_indexes","text":"functions purpose best illustrated example: input: becomes:","code":"list( c(\"A\", \"B\", \"C\"), c(\"A\", \"A\", \"B\"))} list( \"A\" = c(1,2,2), \"B\" = c(1,2), \"C\" = 1 )"},{"path":"/reference/is_absent.html","id":null,"dir":"Reference","previous_headings":"","what":"Is value absent — is_absent","title":"Is value absent — is_absent","text":"Returns true value either NULL, NA \"\". case vector values must NULL/NA/\"\" x regarded absent.","code":""},{"path":"/reference/is_absent.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Is value absent — is_absent","text":"","code":"is_absent(x, na = TRUE, blank = TRUE)"},{"path":"/reference/is_absent.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Is value absent — is_absent","text":"x value check absent na NAs count absent blank blanks .e. \"\" count absent","code":""},{"path":"/reference/is_char_fact.html","id":null,"dir":"Reference","previous_headings":"","what":"Is character or factor — is_char_fact","title":"Is character or factor — is_char_fact","text":"returns true x character factor vector","code":""},{"path":"/reference/is_char_fact.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Is character or factor — is_char_fact","text":"","code":"is_char_fact(x)"},{"path":"/reference/is_char_fact.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Is character or factor — is_char_fact","text":"x character factor vector","code":""},{"path":"/reference/is_char_one.html","id":null,"dir":"Reference","previous_headings":"","what":"Is single character — is_char_one","title":"Is single character — is_char_one","text":"returns true x length 1 character vector","code":""},{"path":"/reference/is_char_one.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Is single character — is_char_one","text":"","code":"is_char_one(x)"},{"path":"/reference/is_char_one.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Is single character — is_char_one","text":"x character vector","code":""},{"path":"/reference/is_in_rbmi_development.html","id":null,"dir":"Reference","previous_headings":"","what":"Is package in development mode? — is_in_rbmi_development","title":"Is package in development mode? — is_in_rbmi_development","text":"Returns TRUE package developed .e. local copy source code actively editing Returns FALSE otherwise","code":""},{"path":"/reference/is_in_rbmi_development.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Is package in development mode? — is_in_rbmi_development","text":"","code":"is_in_rbmi_development()"},{"path":"/reference/is_in_rbmi_development.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Is package in development mode? — is_in_rbmi_development","text":"Main use function parallel processing indicate whether sub-processes need load current development version code whether load main installed package system","code":""},{"path":"/reference/is_num_char_fact.html","id":null,"dir":"Reference","previous_headings":"","what":"Is character, factor or numeric — is_num_char_fact","title":"Is character, factor or numeric — is_num_char_fact","text":"returns true x character, numeric factor vector","code":""},{"path":"/reference/is_num_char_fact.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Is character, factor or numeric — is_num_char_fact","text":"","code":"is_num_char_fact(x)"},{"path":"/reference/is_num_char_fact.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Is character, factor or numeric — is_num_char_fact","text":"x character, numeric factor vector","code":""},{"path":"/reference/locf.html","id":null,"dir":"Reference","previous_headings":"","what":"Last Observation Carried Forward — locf","title":"Last Observation Carried Forward — locf","text":"Returns vector applied last observation carried forward imputation.","code":""},{"path":"/reference/locf.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Last Observation Carried Forward — locf","text":"","code":"locf(x)"},{"path":"/reference/locf.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Last Observation Carried Forward — locf","text":"x vector.","code":""},{"path":"/reference/locf.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Last Observation Carried Forward — locf","text":"","code":"if (FALSE) { locf(c(NA, 1, 2, 3, NA, 4)) # Returns c(NA, 1, 2, 3, 3, 4) }"},{"path":"/reference/longDataConstructor.html","id":null,"dir":"Reference","previous_headings":"","what":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"longdata object allows efficient storage recall longitudinal datasets use bootstrap sampling. object works de-constructing data lists based upon subject id thus enabling efficient lookup.","code":""},{"path":"/reference/longDataConstructor.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"object also handles multiple operations specific rbmi defining whether outcome value MAR / Missing well tracking imputation strategy assigned subject. recognised objects functionality fairly overloaded hoped can split area specific objects / functions future. additions functionality object avoided possible.","code":""},{"path":"/reference/longDataConstructor.html","id":"public-fields","dir":"Reference","previous_headings":"","what":"Public fields","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"data original dataset passed constructor (sorted id visit) vars vars object (list key variables) passed constructor visits character vector containing distinct visit levels ids character vector containing unique ids subject self$data formula formula expressing design matrix data constructed strata numeric vector indicating strata corresponding value self$ids belongs . stratification variable defined default 1 subjects (.e. group). field used part self$sample_ids() function enable stratified bootstrap sampling ice_visit_index list indexed subject storing index number first visit affected ICE. ICE set equal number visits plus 1. values list indexed subject storing numeric vector original (unimputed) outcome values group list indexed subject storing single character indicating imputation group subject belongs defined self$data[id, self$ivars$group] used determine reference group used imputing subjects data. is_mar list indexed subject storing logical values indicating subjects outcome values MAR . list defaulted TRUE subjects & outcomes modified calls self$set_strategies(). Note indicate values missing, variable True outcome values either occurred ICE visit post ICE visit imputation strategy MAR strategies list indexed subject storing single character value indicating imputation strategy assigned subject. list defaulted \"MAR\" subjects modified calls either self$set_strategies() self$update_strategies() strategy_lock list indexed subject storing single logical value indicating whether patients imputation strategy locked . strategy locked means change MAR non-MAR. Strategies can changed non-MAR MAR though trigger warning. Strategies locked patient assigned MAR strategy non-missing ICE date. list populated call self$set_strategies(). indexes list indexed subject storing numeric vector indexes specify rows original dataset belong subject .e. recover full data subject \"pt3\" can use self$data[self$indexes[[\"pt3\"]],]. may seem redundant filtering data directly however enables efficient bootstrap sampling data .e. list populated object initialisation. is_missing list indexed subject storing logical vector indicating whether corresponding outcome subject missing. list populated object initialisation. is_post_ice list indexed subject storing logical vector indicating whether corresponding outcome subject post date ICE. ICE data provided defaults False observations. list populated call self$set_strategies().","code":"indexes <- unlist(self$indexes[c(\"pt3\", \"pt3\")]) self$data[indexes,]"},{"path":[]},{"path":"/reference/longDataConstructor.html","id":"public-methods","dir":"Reference","previous_headings":"","what":"Public methods","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"longDataConstructor$get_data() longDataConstructor$add_subject() longDataConstructor$validate_ids() longDataConstructor$sample_ids() longDataConstructor$extract_by_id() longDataConstructor$update_strategies() longDataConstructor$set_strategies() longDataConstructor$check_has_data_at_each_visit() longDataConstructor$set_strata() longDataConstructor$new() longDataConstructor$clone()","code":""},{"path":"/reference/longDataConstructor.html","id":"method-get-data-","dir":"Reference","previous_headings":"","what":"Method get_data()","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"Returns data.frame based upon required subject IDs. Replaces missing values new ones provided.","code":""},{"path":"/reference/longDataConstructor.html","id":"usage","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"","code":"longDataConstructor$get_data( obj = NULL, nmar.rm = FALSE, na.rm = FALSE, idmap = FALSE )"},{"path":"/reference/longDataConstructor.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"obj Either NULL, character vector subjects IDs imputation list object. See details. nmar.rm Logical value. TRUE remove observations regarded MAR (determined self$is_mar). na.rm Logical value. TRUE remove outcome values missing (determined self$is_missing). idmap Logical value. TRUE add attribute idmap contains mapping new subject ids old subject ids. See details.","code":""},{"path":"/reference/longDataConstructor.html","id":"details-1","dir":"Reference","previous_headings":"","what":"Details","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"obj NULL full original dataset returned. obj character vector new dataset consisting just subjects returned; character vector contains duplicate entries subject returned multiple times. obj imputation_df object (created imputation_df()) subject ids specified object returned missing values filled specified imputation list object. .e. return data.frame consisting observations pt1 twice observations pt3 . first set observations pt1 missing values filled c(1,2,3) second set filled c(4,5,6). length values must equal sum(self$is_missing[[id]]). obj NULL subject IDs scrambled order ensure unique .e. pt2 requested twice process guarantees set observations unique subject ID number. idmap attribute (requested) can used map new ids back old ids.","code":"obj <- imputation_df( imputation_single( id = \"pt1\", values = c(1,2,3)), imputation_single( id = \"pt1\", values = c(4,5,6)), imputation_single( id = \"pt3\", values = c(7,8)) ) longdata$get_data(obj)"},{"path":"/reference/longDataConstructor.html","id":"returns","dir":"Reference","previous_headings":"","what":"Returns","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"data.frame.","code":""},{"path":"/reference/longDataConstructor.html","id":"method-add-subject-","dir":"Reference","previous_headings":"","what":"Method add_subject()","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"function decomposes patient data self$data populates corresponding lists .e. self$is_missing, self$values, self$group, etc. function called upon objects initialization.","code":""},{"path":"/reference/longDataConstructor.html","id":"usage-1","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"","code":"longDataConstructor$add_subject(id)"},{"path":"/reference/longDataConstructor.html","id":"arguments-1","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"id Character subject id exists within self$data.","code":""},{"path":"/reference/longDataConstructor.html","id":"method-validate-ids-","dir":"Reference","previous_headings":"","what":"Method validate_ids()","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"Throws error element ids within source data self$data.","code":""},{"path":"/reference/longDataConstructor.html","id":"usage-2","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"","code":"longDataConstructor$validate_ids(ids)"},{"path":"/reference/longDataConstructor.html","id":"arguments-2","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"ids character vector ids.","code":""},{"path":"/reference/longDataConstructor.html","id":"returns-1","dir":"Reference","previous_headings":"","what":"Returns","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"TRUE","code":""},{"path":"/reference/longDataConstructor.html","id":"method-sample-ids-","dir":"Reference","previous_headings":"","what":"Method sample_ids()","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"Performs random stratified sampling patient ids (replacement) patient equal weight picked within strata (.e dependent many non-missing visits ).","code":""},{"path":"/reference/longDataConstructor.html","id":"usage-3","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"","code":"longDataConstructor$sample_ids()"},{"path":"/reference/longDataConstructor.html","id":"returns-2","dir":"Reference","previous_headings":"","what":"Returns","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"Character vector ids.","code":""},{"path":"/reference/longDataConstructor.html","id":"method-extract-by-id-","dir":"Reference","previous_headings":"","what":"Method extract_by_id()","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"Returns list key information given subject. convenience wrapper save manually grab element.","code":""},{"path":"/reference/longDataConstructor.html","id":"usage-4","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"","code":"longDataConstructor$extract_by_id(id)"},{"path":"/reference/longDataConstructor.html","id":"arguments-3","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"id Character subject id exists within self$data.","code":""},{"path":"/reference/longDataConstructor.html","id":"method-update-strategies-","dir":"Reference","previous_headings":"","what":"Method update_strategies()","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"Convenience function run self$set_strategies(dat_ice, update=TRUE) kept legacy reasons.","code":""},{"path":"/reference/longDataConstructor.html","id":"usage-5","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"","code":"longDataConstructor$update_strategies(dat_ice)"},{"path":"/reference/longDataConstructor.html","id":"arguments-4","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"dat_ice data.frame containing ICE information see impute() format dataframe.","code":""},{"path":"/reference/longDataConstructor.html","id":"method-set-strategies-","dir":"Reference","previous_headings":"","what":"Method set_strategies()","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"Updates self$strategies, self$is_mar, self$is_post_ice variables based upon provided ICE information.","code":""},{"path":"/reference/longDataConstructor.html","id":"usage-6","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"","code":"longDataConstructor$set_strategies(dat_ice = NULL, update = FALSE)"},{"path":"/reference/longDataConstructor.html","id":"arguments-5","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"dat_ice data.frame containing ICE information. See details. update Logical, indicates ICE data used update. See details.","code":""},{"path":"/reference/longDataConstructor.html","id":"details-2","dir":"Reference","previous_headings":"","what":"Details","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"See draws() specification dat_ice update=FALSE. See impute() format dat_ice update=TRUE. update=TRUE function ensures MAR strategies changed non-MAR presence post-ICE observations.","code":""},{"path":"/reference/longDataConstructor.html","id":"method-check-has-data-at-each-visit-","dir":"Reference","previous_headings":"","what":"Method check_has_data_at_each_visit()","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"Ensures visits least 1 observed \"MAR\" observation. Throws error criteria met. ensure initial MMRM can resolved.","code":""},{"path":"/reference/longDataConstructor.html","id":"usage-7","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"","code":"longDataConstructor$check_has_data_at_each_visit()"},{"path":"/reference/longDataConstructor.html","id":"method-set-strata-","dir":"Reference","previous_headings":"","what":"Method set_strata()","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"Populates self$strata variable. user specified stratification variables first visit used determine value variables. stratification variables specified everyone defined strata 1.","code":""},{"path":"/reference/longDataConstructor.html","id":"usage-8","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"","code":"longDataConstructor$set_strata()"},{"path":"/reference/longDataConstructor.html","id":"method-new-","dir":"Reference","previous_headings":"","what":"Method new()","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"Constructor function.","code":""},{"path":"/reference/longDataConstructor.html","id":"usage-9","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"","code":"longDataConstructor$new(data, vars)"},{"path":"/reference/longDataConstructor.html","id":"arguments-6","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"data longitudinal dataset. vars ivars object created set_vars().","code":""},{"path":"/reference/longDataConstructor.html","id":"method-clone-","dir":"Reference","previous_headings":"","what":"Method clone()","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"objects class cloneable method.","code":""},{"path":"/reference/longDataConstructor.html","id":"usage-10","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"","code":"longDataConstructor$clone(deep = FALSE)"},{"path":"/reference/longDataConstructor.html","id":"arguments-7","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"deep Whether make deep clone.","code":""},{"path":"/reference/ls_design.html","id":null,"dir":"Reference","previous_headings":"","what":"Calculate design vector for the lsmeans — ls_design","title":"Calculate design vector for the lsmeans — ls_design","text":"Calculates design vector required generate lsmean standard error. ls_design_equal calculates applying equal weight per covariate combination whilst ls_design_proportional applies weighting proportional frequency covariate combination occurred actual dataset.","code":""},{"path":"/reference/ls_design.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Calculate design vector for the lsmeans — ls_design","text":"","code":"ls_design_equal(data, frm, covars, fix) ls_design_proportional(data, frm, covars, fix)"},{"path":"/reference/ls_design.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Calculate design vector for the lsmeans — ls_design","text":"data data.frame frm Formula used fit original model covars character vector variables names exist data extracted (ls_design_equal ) fix named list variables fixed values","code":""},{"path":"/reference/lsmeans.html","id":null,"dir":"Reference","previous_headings":"","what":"Least Square Means — lsmeans","title":"Least Square Means — lsmeans","text":"Estimates least square means linear model. done generating prediction model using hypothetical observation constructed averaging data. See details information.","code":""},{"path":"/reference/lsmeans.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Least Square Means — lsmeans","text":"","code":"lsmeans(model, ..., .weights = c(\"proportional\", \"equal\"))"},{"path":"/reference/lsmeans.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Least Square Means — lsmeans","text":"model model created lm. ... Fixes specific variables specific values .e. trt = 1 age = 50. name argument must name variable within dataset. .weights Character, specifies whether use \"proportional\" \"equal\" weighting categorical covariate combination calculating lsmeans.","code":""},{"path":"/reference/lsmeans.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Least Square Means — lsmeans","text":"lsmeans obtained calculating hypothetical patients predicting expected values. hypothetical patients constructed expanding possible combinations categorical covariate setting numerical covariates equal mean. final lsmean value calculated averaging hypothetical patients. .weights equals \"proportional\" values weighted frequency occur full dataset. .weights equals \"equal\" hypothetical patient given equal weight regardless actually occurs dataset. Use ... argument fix specific variables specific values. See references identical implementations done SAS R via emmeans package. function attempts re-implement emmeans derivation standard linear models without include dependencies.","code":""},{"path":"/reference/lsmeans.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Least Square Means — lsmeans","text":"https://CRAN.R-project.org/package=emmeans https://documentation.sas.com/doc/en/pgmsascdc/9.4_3.3/statug/statug_glm_details41.htm","code":""},{"path":"/reference/lsmeans.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Least Square Means — lsmeans","text":"","code":"if (FALSE) { mod <- lm(Sepal.Length ~ Species + Petal.Length, data = iris) lsmeans(mod) lsmeans(mod, Species = \"virginica\") lsmeans(mod, Species = \"versicolor\") lsmeans(mod, Species = \"versicolor\", Petal.Length = 1) }"},{"path":"/reference/method.html","id":null,"dir":"Reference","previous_headings":"","what":"Set the multiple imputation methodology — method","title":"Set the multiple imputation methodology — method","text":"functions determine methods rbmi use creating imputation models, generating imputed values pooling results.","code":""},{"path":"/reference/method.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Set the multiple imputation methodology — method","text":"","code":"method_bayes( burn_in = 200, burn_between = 50, same_cov = TRUE, n_samples = 20, seed = sample.int(.Machine$integer.max, 1) ) method_approxbayes( covariance = c(\"us\", \"toep\", \"cs\", \"ar1\"), threshold = 0.01, same_cov = TRUE, REML = TRUE, n_samples = 20 ) method_condmean( covariance = c(\"us\", \"toep\", \"cs\", \"ar1\"), threshold = 0.01, same_cov = TRUE, REML = TRUE, n_samples = NULL, type = c(\"bootstrap\", \"jackknife\") ) method_bmlmi( covariance = c(\"us\", \"toep\", \"cs\", \"ar1\"), threshold = 0.01, same_cov = TRUE, REML = TRUE, B = 20, D = 2 )"},{"path":"/reference/method.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Set the multiple imputation methodology — method","text":"burn_in numeric specifies many observations discarded prior extracting actual samples. Note sampler initialized maximum likelihood estimates weakly informative prior used thus theory value need high. burn_between numeric specifies \"thinning\" rate .e. many observations discarded sample. used prevent issues associated autocorrelation samples. same_cov logical, TRUE imputation model fitted using single shared covariance matrix observations. FALSE separate covariance matrix fit group determined group argument set_vars(). n_samples numeric determines many imputed datasets generated. case method_condmean(type = \"jackknife\") argument must set NULL. See details. seed numeric specifies seed used call Stan. argument passed onto seed argument rstan::sampling(). Note required method_bayes(), methods can achieve reproducible results setting seed via set.seed(). See details. covariance character string specifies structure covariance matrix used imputation model. Must one \"us\" (default), \"toep\", \"cs\" \"ar1\". See details. threshold numeric 0 1, specifies proportion bootstrap datasets can fail produce valid samples error thrown. See details. REML logical indicating whether use REML estimation rather maximum likelihood. type character string specifies resampling method used perform inference conditional mean imputation approach (set via method_condmean()) used. Must one \"bootstrap\" \"jackknife\". B numeric determines number bootstrap samples method_bmlmi. D numeric determines number random imputations bootstrap sample. Needed method_bmlmi().","code":""},{"path":"/reference/method.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Set the multiple imputation methodology — method","text":"case method_condmean(type = \"bootstrap\") n_samples + 1 imputation models datasets generated first sample based original dataset whilst n_samples samples bootstrapped datasets. Likewise, method_condmean(type = \"jackknife\") length(unique(data$subjid)) + 1 imputation models datasets generated. cases represented n + 1 displayed print message. user able specify different covariance structures using covariance argument. Currently supported structures include: Unstructured (\"us\") Toeplitz (\"toep\") Compound Symmetry (\"cs\") Autoregression-1 (\"ar1\") Note present Bayesian methods support unstructured. case method_condmean(type = \"bootstrap\"), method_approxbayes() method_bmlmi() repeated bootstrap samples original dataset taken MMRM fitted sample. Due randomness sampled datasets, well limitations optimisers used fit models, uncommon estimates particular dataset generated. instances rbmi designed throw bootstrapped dataset try another. However ensure errors due chance due underlying misspecification data /model tolerance limit set many samples can discarded. tolerance limit reached error thrown process aborted. tolerance limit defined ceiling(threshold * n_samples). Note jackknife method estimates need generated leave-one-datasets error thrown fail fit. Please note time writing (September 2021) Stan unable produce reproducible samples across different operating systems even seed used. care must taken using Stan across different machines. information limitation please consult Stan documentation https://mc-stan.org/docs/2_27/reference-manual/reproducibility-chapter.html","code":""},{"path":"/reference/parametric_ci.html","id":null,"dir":"Reference","previous_headings":"","what":"Calculate parametric confidence intervals — parametric_ci","title":"Calculate parametric confidence intervals — parametric_ci","text":"Calculates confidence intervals based upon parametric distribution.","code":""},{"path":"/reference/parametric_ci.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Calculate parametric confidence intervals — parametric_ci","text":"","code":"parametric_ci(point, se, alpha, alternative, qfun, pfun, ...)"},{"path":"/reference/parametric_ci.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Calculate parametric confidence intervals — parametric_ci","text":"point point estimate. se standard error point estimate. using non-\"normal\" distribution set 1. alpha type 1 error rate, value 0 1. alternative character string specifying alternative hypothesis, must one \"two.sided\" (default), \"greater\" \"less\". qfun quantile function assumed distribution .e. qnorm. pfun CDF function assumed distribution .e. pnorm. ... additional arguments passed qfun pfun .e. df = 102.","code":""},{"path":"/reference/pool.html","id":null,"dir":"Reference","previous_headings":"","what":"Pool analysis results obtained from the imputed datasets — pool","title":"Pool analysis results obtained from the imputed datasets — pool","text":"Pool analysis results obtained imputed datasets","code":""},{"path":"/reference/pool.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Pool analysis results obtained from the imputed datasets — pool","text":"","code":"pool( results, conf.level = 0.95, alternative = c(\"two.sided\", \"less\", \"greater\"), type = c(\"percentile\", \"normal\") ) # S3 method for pool as.data.frame(x, ...) # S3 method for pool print(x, ...)"},{"path":"/reference/pool.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Pool analysis results obtained from the imputed datasets — pool","text":"results analysis object created analyse(). conf.level confidence level returned confidence interval. Must single number 0 1. Default 0.95. alternative character string specifying alternative hypothesis, must one \"two.sided\" (default), \"greater\" \"less\". type character string either \"percentile\" (default) \"normal\". Determines method used calculate bootstrap confidence intervals. See details. used method_condmean(type = \"bootstrap\") specified original call draws(). x pool object generated pool(). ... used.","code":""},{"path":"/reference/pool.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Pool analysis results obtained from the imputed datasets — pool","text":"calculation used generate point estimate, standard errors confidence interval depends upon method specified original call draws(); particular: method_approxbayes() & method_bayes() use Rubin's rules pool estimates variances across multiple imputed datasets, Barnard-Rubin rule pool degree's freedom; see Little & Rubin (2002). method_condmean(type = \"bootstrap\") uses percentile normal approximation; see Efron & Tibshirani (1994). Note percentile bootstrap, standard error calculated, .e. standard errors NA object / data.frame. method_condmean(type = \"jackknife\") uses standard jackknife variance formula; see Efron & Tibshirani (1994). method_bmlmi uses pooling procedure Bootstrapped Maximum Likelihood MI (BMLMI). See Von Hippel & Bartlett (2021).","code":""},{"path":"/reference/pool.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Pool analysis results obtained from the imputed datasets — pool","text":"Bradley Efron Robert J Tibshirani. introduction bootstrap. CRC press, 1994. [Section 11] Roderick J. . Little Donald B. Rubin. Statistical Analysis Missing Data, Second Edition. John Wiley & Sons, Hoboken, New Jersey, 2002. [Section 5.4] Von Hippel, Paul T Bartlett, Jonathan W. Maximum likelihood multiple imputation: Faster imputations consistent standard errors without posterior draws. 2021.","code":""},{"path":"/reference/pool_bootstrap_normal.html","id":null,"dir":"Reference","previous_headings":"","what":"Bootstrap Pooling via normal approximation — pool_bootstrap_normal","title":"Bootstrap Pooling via normal approximation — pool_bootstrap_normal","text":"Get point estimate, confidence interval p-value using normal approximation.","code":""},{"path":"/reference/pool_bootstrap_normal.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Bootstrap Pooling via normal approximation — pool_bootstrap_normal","text":"","code":"pool_bootstrap_normal(est, conf.level, alternative)"},{"path":"/reference/pool_bootstrap_normal.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Bootstrap Pooling via normal approximation — pool_bootstrap_normal","text":"est numeric vector point estimates bootstrap sample. conf.level confidence level returned confidence interval. Must single number 0 1. Default 0.95. alternative character string specifying alternative hypothesis, must one \"two.sided\" (default), \"greater\" \"less\".","code":""},{"path":"/reference/pool_bootstrap_normal.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Bootstrap Pooling via normal approximation — pool_bootstrap_normal","text":"point estimate taken first element est. remaining n-1 values est used generate confidence intervals.","code":""},{"path":"/reference/pool_bootstrap_percentile.html","id":null,"dir":"Reference","previous_headings":"","what":"Bootstrap Pooling via Percentiles — pool_bootstrap_percentile","title":"Bootstrap Pooling via Percentiles — pool_bootstrap_percentile","text":"Get point estimate, confidence interval p-value using percentiles. Note quantile \"type=6\" used, see stats::quantile() details.","code":""},{"path":"/reference/pool_bootstrap_percentile.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Bootstrap Pooling via Percentiles — pool_bootstrap_percentile","text":"","code":"pool_bootstrap_percentile(est, conf.level, alternative)"},{"path":"/reference/pool_bootstrap_percentile.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Bootstrap Pooling via Percentiles — pool_bootstrap_percentile","text":"est numeric vector point estimates bootstrap sample. conf.level confidence level returned confidence interval. Must single number 0 1. Default 0.95. alternative character string specifying alternative hypothesis, must one \"two.sided\" (default), \"greater\" \"less\".","code":""},{"path":"/reference/pool_bootstrap_percentile.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Bootstrap Pooling via Percentiles — pool_bootstrap_percentile","text":"point estimate taken first element est. remaining n-1 values est used generate confidence intervals.","code":""},{"path":"/reference/pool_internal.html","id":null,"dir":"Reference","previous_headings":"","what":"Internal Pool Methods — pool_internal","title":"Internal Pool Methods — pool_internal","text":"Dispatches pool methods based upon results object class. See pool() details.","code":""},{"path":"/reference/pool_internal.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Internal Pool Methods — pool_internal","text":"","code":"pool_internal(results, conf.level, alternative, type, D) # S3 method for jackknife pool_internal(results, conf.level, alternative, type, D) # S3 method for bootstrap pool_internal( results, conf.level, alternative, type = c(\"percentile\", \"normal\"), D ) # S3 method for bmlmi pool_internal(results, conf.level, alternative, type, D) # S3 method for rubin pool_internal(results, conf.level, alternative, type, D)"},{"path":"/reference/pool_internal.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Internal Pool Methods — pool_internal","text":"results list results .e. x$results element analyse object created analyse()). conf.level confidence level returned confidence interval. Must single number 0 1. Default 0.95. alternative character string specifying alternative hypothesis, must one \"two.sided\" (default), \"greater\" \"less\". type character string either \"percentile\" (default) \"normal\". Determines method used calculate bootstrap confidence intervals. See details. used method_condmean(type = \"bootstrap\") specified original call draws(). D numeric representing number imputations bootstrap sample BMLMI method.","code":""},{"path":"/reference/prepare_stan_data.html","id":null,"dir":"Reference","previous_headings":"","what":"Prepare input data to run the Stan model — prepare_stan_data","title":"Prepare input data to run the Stan model — prepare_stan_data","text":"Prepare input data run Stan model. Creates / calculates required inputs required data{} block MMRM Stan program.","code":""},{"path":"/reference/prepare_stan_data.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Prepare input data to run the Stan model — prepare_stan_data","text":"","code":"prepare_stan_data(ddat, subjid, visit, outcome, group)"},{"path":"/reference/prepare_stan_data.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Prepare input data to run the Stan model — prepare_stan_data","text":"ddat design matrix subjid Character vector containing subjects IDs. visit Vector containing visits. outcome Numeric vector containing outcome variable. group Vector containing group variable.","code":""},{"path":"/reference/prepare_stan_data.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Prepare input data to run the Stan model — prepare_stan_data","text":"stan_data object. named list per data{} block related Stan file. particular returns: N - number rows design matrix P - number columns design matrix G - number distinct covariance matrix groups (.e. length(unique(group))) n_visit - number unique outcome visits n_pat - total number pattern groups (defined missingness patterns & covariance group) pat_G - Index Sigma pattern group use pat_n_pt - number patients within pattern group pat_n_visit - number non-missing visits pattern group pat_sigma_index - rows/cols Sigma subset pattern group (padded 0's) y - outcome variable Q - design matrix (QR decomposition) R - R matrix QR decomposition design matrix","code":""},{"path":"/reference/prepare_stan_data.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Prepare input data to run the Stan model — prepare_stan_data","text":"group argument determines covariance matrix group subject belongs . want subjects use shared covariance matrix set group \"1\" everyone.","code":""},{"path":"/reference/print.analysis.html","id":null,"dir":"Reference","previous_headings":"","what":"Print analysis object — print.analysis","title":"Print analysis object — print.analysis","text":"Print analysis object","code":""},{"path":"/reference/print.analysis.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Print analysis object — print.analysis","text":"","code":"# S3 method for analysis print(x, ...)"},{"path":"/reference/print.analysis.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Print analysis object — print.analysis","text":"x analysis object generated analyse(). ... used.","code":""},{"path":"/reference/print.draws.html","id":null,"dir":"Reference","previous_headings":"","what":"Print draws object — print.draws","title":"Print draws object — print.draws","text":"Print draws object","code":""},{"path":"/reference/print.draws.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Print draws object — print.draws","text":"","code":"# S3 method for draws print(x, ...)"},{"path":"/reference/print.draws.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Print draws object — print.draws","text":"x draws object generated draws(). ... used.","code":""},{"path":"/reference/print.imputation.html","id":null,"dir":"Reference","previous_headings":"","what":"Print imputation object — print.imputation","title":"Print imputation object — print.imputation","text":"Print imputation object","code":""},{"path":"/reference/print.imputation.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Print imputation object — print.imputation","text":"","code":"# S3 method for imputation print(x, ...)"},{"path":"/reference/print.imputation.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Print imputation object — print.imputation","text":"x imputation object generated impute(). ... used.","code":""},{"path":"/reference/progressLogger.html","id":null,"dir":"Reference","previous_headings":"","what":"R6 Class for printing current sampling progress — progressLogger","title":"R6 Class for printing current sampling progress — progressLogger","text":"Object initalised total number iterations expected occur. User can update object add method indicate many iterations just occurred. Every time step * 100 % iterations occurred message printed console. Use quiet argument prevent object printing anything ","code":""},{"path":"/reference/progressLogger.html","id":"public-fields","dir":"Reference","previous_headings":"","what":"Public fields","title":"R6 Class for printing current sampling progress — progressLogger","text":"step real, percentage iterations allow printing progress console step_current integer, total number iterations completed since progress last printed console n integer, current number completed iterations n_max integer, total number expected iterations completed acts denominator calculating progress percentages quiet logical holds whether print anything","code":""},{"path":[]},{"path":"/reference/progressLogger.html","id":"public-methods","dir":"Reference","previous_headings":"","what":"Public methods","title":"R6 Class for printing current sampling progress — progressLogger","text":"progressLogger$new() progressLogger$add() progressLogger$print_progress() progressLogger$clone()","code":""},{"path":"/reference/progressLogger.html","id":"method-new-","dir":"Reference","previous_headings":"","what":"Method new()","title":"R6 Class for printing current sampling progress — progressLogger","text":"Create progressLogger object","code":""},{"path":"/reference/progressLogger.html","id":"usage","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for printing current sampling progress — progressLogger","text":"","code":"progressLogger$new(n_max, quiet = FALSE, step = 0.1)"},{"path":"/reference/progressLogger.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for printing current sampling progress — progressLogger","text":"n_max integer, sets field n_max quiet logical, sets field quiet step real, sets field step","code":""},{"path":"/reference/progressLogger.html","id":"method-add-","dir":"Reference","previous_headings":"","what":"Method add()","title":"R6 Class for printing current sampling progress — progressLogger","text":"Records n iterations completed add number current step count (step_current) print progress message log step limit (step) reached. function nothing quiet set TRUE","code":""},{"path":"/reference/progressLogger.html","id":"usage-1","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for printing current sampling progress — progressLogger","text":"","code":"progressLogger$add(n)"},{"path":"/reference/progressLogger.html","id":"arguments-1","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for printing current sampling progress — progressLogger","text":"n number successfully complete iterations since add() last called","code":""},{"path":"/reference/progressLogger.html","id":"method-print-progress-","dir":"Reference","previous_headings":"","what":"Method print_progress()","title":"R6 Class for printing current sampling progress — progressLogger","text":"method print current state progress","code":""},{"path":"/reference/progressLogger.html","id":"usage-2","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for printing current sampling progress — progressLogger","text":"","code":"progressLogger$print_progress()"},{"path":"/reference/progressLogger.html","id":"method-clone-","dir":"Reference","previous_headings":"","what":"Method clone()","title":"R6 Class for printing current sampling progress — progressLogger","text":"objects class cloneable method.","code":""},{"path":"/reference/progressLogger.html","id":"usage-3","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for printing current sampling progress — progressLogger","text":"","code":"progressLogger$clone(deep = FALSE)"},{"path":"/reference/progressLogger.html","id":"arguments-2","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for printing current sampling progress — progressLogger","text":"deep Whether make deep clone.","code":""},{"path":"/reference/pval_percentile.html","id":null,"dir":"Reference","previous_headings":"","what":"P-value of percentile bootstrap — pval_percentile","title":"P-value of percentile bootstrap — pval_percentile","text":"Determines (necessarily unique) quantile (type=6) \"est\" gives value 0 , derive p-value corresponding percentile bootstrap via inversion.","code":""},{"path":"/reference/pval_percentile.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"P-value of percentile bootstrap — pval_percentile","text":"","code":"pval_percentile(est)"},{"path":"/reference/pval_percentile.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"P-value of percentile bootstrap — pval_percentile","text":"est numeric vector point estimates bootstrap sample.","code":""},{"path":"/reference/pval_percentile.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"P-value of percentile bootstrap — pval_percentile","text":"named numeric vector length 2 containing p-value H_0: theta=0 vs H_A: theta>0 (\"pval_greater\") p-value H_0: theta=0 vs H_A: theta<0 (\"pval_less\").","code":""},{"path":"/reference/pval_percentile.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"P-value of percentile bootstrap — pval_percentile","text":"p-value H_0: theta=0 vs H_A: theta>0 value alpha q_alpha = 0. least one estimate equal zero returns largest alpha q_alpha = 0. bootstrap estimates > 0 returns 0; bootstrap estimates < 0 returns 1. Analogous reasoning applied p-value H_0: theta=0 vs H_A: theta<0.","code":""},{"path":"/reference/random_effects_expr.html","id":null,"dir":"Reference","previous_headings":"","what":"Construct random effects formula — random_effects_expr","title":"Construct random effects formula — random_effects_expr","text":"Constructs character representation random effects formula fitting MMRM subject visit format required mmrm::mmrm().","code":""},{"path":"/reference/random_effects_expr.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Construct random effects formula — random_effects_expr","text":"","code":"random_effects_expr( cov_struct = c(\"us\", \"toep\", \"cs\", \"ar1\"), cov_by_group = FALSE )"},{"path":"/reference/random_effects_expr.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Construct random effects formula — random_effects_expr","text":"cov_struct Character - covariance structure used, must one \"us\", \"toep\", \"cs\", \"ar1\" cov_by_group Boolean - Whenever use separate covariances per group level","code":""},{"path":"/reference/random_effects_expr.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Construct random effects formula — random_effects_expr","text":"example assuming user specified covariance structure \"us\" groups provided return cov_by_group set FALSE indicates separate covariance matrices required per group following returned:","code":"us(visit | subjid) us( visit | group / subjid )"},{"path":"/reference/rbmi-package.html","id":null,"dir":"Reference","previous_headings":"","what":"rbmi: Reference Based Multiple Imputation — rbmi-package","title":"rbmi: Reference Based Multiple Imputation — rbmi-package","text":"rbmi package used perform reference based multiple imputation. package provides implementations common, patient-specific imputation strategies whilst allowing user select various standard Bayesian frequentist approaches. package designed around 4 core functions: draws() - Fits multiple imputation models impute() - Imputes multiple datasets analyse() - Analyses multiple datasets pool() - Pools multiple results single statistic learn rbmi, please see quickstart vignette: vignette(topic= \"quickstart\", package = \"rbmi\")","code":""},{"path":[]},{"path":"/reference/rbmi-package.html","id":"author","dir":"Reference","previous_headings":"","what":"Author","title":"rbmi: Reference Based Multiple Imputation — rbmi-package","text":"Maintainer: Craig Gower-Page craig.gower-page@roche.com Authors: Alessandro Noci alessandro.noci@roche.com contributors: Marcel Wolbers marcel.wolbers@roche.com [contributor] Roche [copyright holder, funder]","code":""},{"path":"/reference/record.html","id":null,"dir":"Reference","previous_headings":"","what":"Capture all Output — record","title":"Capture all Output — record","text":"function silences warnings, errors & messages instead returns list containing results (error) + warning error messages character vectors.","code":""},{"path":"/reference/record.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Capture all Output — record","text":"","code":"record(expr)"},{"path":"/reference/record.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Capture all Output — record","text":"expr expression executed","code":""},{"path":"/reference/record.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Capture all Output — record","text":"list containing results - object returned expr list() error thrown warnings - NULL character vector warnings thrown errors - NULL string error thrown messages - NULL character vector messages produced","code":""},{"path":"/reference/record.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Capture all Output — record","text":"","code":"if (FALSE) { record({ x <- 1 y <- 2 warning(\"something went wrong\") message(\"O nearly done\") x + y }) }"},{"path":"/reference/recursive_reduce.html","id":null,"dir":"Reference","previous_headings":"","what":"recursive_reduce — recursive_reduce","title":"recursive_reduce — recursive_reduce","text":"Utility function used replicated purrr::reduce. Recursively applies function list elements 1 element remains","code":""},{"path":"/reference/recursive_reduce.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"recursive_reduce — recursive_reduce","text":"","code":"recursive_reduce(.l, .f)"},{"path":"/reference/recursive_reduce.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"recursive_reduce — recursive_reduce","text":".l list values apply function .f function apply element list turn .e. .l[[1]] <- .f( .l[[1]] , .l[[2]]) ; .l[[1]] <- .f( .l[[1]] , .l[[3]])","code":""},{"path":"/reference/remove_if_all_missing.html","id":null,"dir":"Reference","previous_headings":"","what":"Remove subjects from dataset if they have no observed values — remove_if_all_missing","title":"Remove subjects from dataset if they have no observed values — remove_if_all_missing","text":"function takes data.frame variables visit, outcome & subjid. removes rows given subjid non-missing values outcome.","code":""},{"path":"/reference/remove_if_all_missing.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Remove subjects from dataset if they have no observed values — remove_if_all_missing","text":"","code":"remove_if_all_missing(dat)"},{"path":"/reference/remove_if_all_missing.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Remove subjects from dataset if they have no observed values — remove_if_all_missing","text":"dat data.frame","code":""},{"path":"/reference/rubin_df.html","id":null,"dir":"Reference","previous_headings":"","what":"Barnard and Rubin degrees of freedom adjustment — rubin_df","title":"Barnard and Rubin degrees of freedom adjustment — rubin_df","text":"Compute degrees freedom according Barnard-Rubin formula.","code":""},{"path":"/reference/rubin_df.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Barnard and Rubin degrees of freedom adjustment — rubin_df","text":"","code":"rubin_df(v_com, var_b, var_t, M)"},{"path":"/reference/rubin_df.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Barnard and Rubin degrees of freedom adjustment — rubin_df","text":"v_com Positive number representing degrees freedom complete-data analysis. var_b -variance point estimate across multiply imputed datasets. var_t Total-variance point estimate according Rubin's rules. M Number imputations.","code":""},{"path":"/reference/rubin_df.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Barnard and Rubin degrees of freedom adjustment — rubin_df","text":"Degrees freedom according Barnard-Rubin formula. See Barnard-Rubin (1999).","code":""},{"path":"/reference/rubin_df.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Barnard and Rubin degrees of freedom adjustment — rubin_df","text":"computation takes account limit cases missing data (.e. -variance var_b zero) complete-data degrees freedom set Inf. Moreover, v_com given NA, function returns Inf.","code":""},{"path":"/reference/rubin_df.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Barnard and Rubin degrees of freedom adjustment — rubin_df","text":"Barnard, J. Rubin, D.B. (1999). Small sample degrees freedom multiple imputation. Biometrika, 86, 948-955.","code":""},{"path":"/reference/rubin_rules.html","id":null,"dir":"Reference","previous_headings":"","what":"Combine estimates using Rubin's rules — rubin_rules","title":"Combine estimates using Rubin's rules — rubin_rules","text":"Pool together results M complete-data analyses according Rubin's rules. See details.","code":""},{"path":"/reference/rubin_rules.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Combine estimates using Rubin's rules — rubin_rules","text":"","code":"rubin_rules(ests, ses, v_com)"},{"path":"/reference/rubin_rules.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Combine estimates using Rubin's rules — rubin_rules","text":"ests Numeric vector containing point estimates complete-data analyses. ses Numeric vector containing standard errors complete-data analyses. v_com Positive number representing degrees freedom complete-data analysis.","code":""},{"path":"/reference/rubin_rules.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Combine estimates using Rubin's rules — rubin_rules","text":"list containing: est_point: pooled point estimate according Little-Rubin (2002). var_t: total variance according Little-Rubin (2002). df: degrees freedom according Barnard-Rubin (1999).","code":""},{"path":"/reference/rubin_rules.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Combine estimates using Rubin's rules — rubin_rules","text":"rubin_rules applies Rubin's rules (Rubin, 1987) pooling together results multiple imputation procedure. pooled point estimate est_point average across point estimates complete-data analyses (given input argument ests). total variance var_t sum two terms representing within-variance -variance (see Little-Rubin (2002)). function also returns df, estimated pooled degrees freedom according Barnard-Rubin (1999) can used inference based t-distribution.","code":""},{"path":"/reference/rubin_rules.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Combine estimates using Rubin's rules — rubin_rules","text":"Barnard, J. Rubin, D.B. (1999). Small sample degrees freedom multiple imputation. Biometrika, 86, 948-955 Roderick J. . Little Donald B. Rubin. Statistical Analysis Missing Data, Second Edition. John Wiley & Sons, Hoboken, New Jersey, 2002. [Section 5.4]","code":""},{"path":[]},{"path":"/reference/sample_ids.html","id":null,"dir":"Reference","previous_headings":"","what":"Sample Patient Ids — sample_ids","title":"Sample Patient Ids — sample_ids","text":"Performs stratified bootstrap sample IDS ensuring return vector length input vector","code":""},{"path":"/reference/sample_ids.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Sample Patient Ids — sample_ids","text":"","code":"sample_ids(ids, strata = rep(1, length(ids)))"},{"path":"/reference/sample_ids.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Sample Patient Ids — sample_ids","text":"ids vector sample strata strata indicator, ids sampled within strata ensuring numbers strata maintained","code":""},{"path":"/reference/sample_ids.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Sample Patient Ids — sample_ids","text":"","code":"if (FALSE) { sample_ids( c(\"a\", \"b\", \"c\", \"d\"), strata = c(1,1,2,2)) }"},{"path":"/reference/sample_list.html","id":null,"dir":"Reference","previous_headings":"","what":"Create and validate a sample_list object — sample_list","title":"Create and validate a sample_list object — sample_list","text":"Given list sample_single objects generate sample_single(), creates sample_list objects validate .","code":""},{"path":"/reference/sample_list.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create and validate a sample_list object — sample_list","text":"","code":"sample_list(...)"},{"path":"/reference/sample_list.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create and validate a sample_list object — sample_list","text":"... list sample_single objects.","code":""},{"path":"/reference/sample_mvnorm.html","id":null,"dir":"Reference","previous_headings":"","what":"Sample random values from the multivariate normal distribution — sample_mvnorm","title":"Sample random values from the multivariate normal distribution — sample_mvnorm","text":"Sample random values multivariate normal distribution","code":""},{"path":"/reference/sample_mvnorm.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Sample random values from the multivariate normal distribution — sample_mvnorm","text":"","code":"sample_mvnorm(mu, sigma)"},{"path":"/reference/sample_mvnorm.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Sample random values from the multivariate normal distribution — sample_mvnorm","text":"mu mean vector sigma covariance matrix Samples multivariate normal variables multiplying univariate random normal variables cholesky decomposition covariance matrix. mu length 1 just uses rnorm instead.","code":""},{"path":"/reference/sample_single.html","id":null,"dir":"Reference","previous_headings":"","what":"Create object of sample_single class — sample_single","title":"Create object of sample_single class — sample_single","text":"Creates object class sample_single named list containing input parameters validate .","code":""},{"path":"/reference/sample_single.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create object of sample_single class — sample_single","text":"","code":"sample_single( ids, beta = NA, sigma = NA, theta = NA, failed = any(is.na(beta)), ids_samp = ids )"},{"path":"/reference/sample_single.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create object of sample_single class — sample_single","text":"ids Vector characters containing ids subjects included original dataset. beta Numeric vector estimated regression coefficients. sigma List estimated covariance matrices (one level vars$group). theta Numeric vector transformed covariances. failed Logical. TRUE model fit failed. ids_samp Vector characters containing ids subjects included given sample.","code":""},{"path":"/reference/sample_single.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Create object of sample_single class — sample_single","text":"named list class sample_single. contains following: ids vector characters containing ids subjects included original dataset. beta numeric vector estimated regression coefficients. sigma list estimated covariance matrices (one level vars$group). theta numeric vector transformed covariances. failed logical. TRUE model fit failed. ids_samp vector characters containing ids subjects included given sample.","code":""},{"path":"/reference/scalerConstructor.html","id":null,"dir":"Reference","previous_headings":"","what":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"Scales design matrix non-categorical columns mean 0 standard deviation 1.","code":""},{"path":"/reference/scalerConstructor.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"object initialisation used determine relevant mean SD's scale scaling (un-scaling) performed relevant object methods. Un-scaling done linear model Beta Sigma coefficients. purpose first column dataset scaled assumed outcome variable variables assumed post-transformation predictor variables (.e. dummy variables already expanded).","code":""},{"path":"/reference/scalerConstructor.html","id":"public-fields","dir":"Reference","previous_headings":"","what":"Public fields","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"centre Vector column means. first value outcome variable, variables predictors. scales Vector column standard deviations. first value outcome variable, variables predictors.","code":""},{"path":[]},{"path":"/reference/scalerConstructor.html","id":"public-methods","dir":"Reference","previous_headings":"","what":"Public methods","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"scalerConstructor$new() scalerConstructor$scale() scalerConstructor$unscale_sigma() scalerConstructor$unscale_beta() scalerConstructor$clone()","code":""},{"path":"/reference/scalerConstructor.html","id":"method-new-","dir":"Reference","previous_headings":"","what":"Method new()","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"Uses dat determine relevant column means standard deviations use scaling un-scaling future datasets. Implicitly assumes new datasets column order dat","code":""},{"path":"/reference/scalerConstructor.html","id":"usage","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"","code":"scalerConstructor$new(dat)"},{"path":"/reference/scalerConstructor.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"dat data.frame matrix. columns must numeric (.e dummy variables, must already expanded ).","code":""},{"path":"/reference/scalerConstructor.html","id":"details-1","dir":"Reference","previous_headings":"","what":"Details","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"Categorical columns (determined values entirely 1 0) scaled. achieved setting corresponding values centre 0 scale 1.","code":""},{"path":"/reference/scalerConstructor.html","id":"method-scale-","dir":"Reference","previous_headings":"","what":"Method scale()","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"Scales dataset continuous variables mean 0 standard deviation 1.","code":""},{"path":"/reference/scalerConstructor.html","id":"usage-1","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"","code":"scalerConstructor$scale(dat)"},{"path":"/reference/scalerConstructor.html","id":"arguments-1","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"dat data.frame matrix whose columns numeric (.e. dummy variables expanded ) whose columns order dataset used initialization function.","code":""},{"path":"/reference/scalerConstructor.html","id":"method-unscale-sigma-","dir":"Reference","previous_headings":"","what":"Method unscale_sigma()","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"Unscales sigma value (matrix) estimated linear model using design matrix scaled object. function works first column initialisation data.frame outcome variable.","code":""},{"path":"/reference/scalerConstructor.html","id":"usage-2","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"","code":"scalerConstructor$unscale_sigma(sigma)"},{"path":"/reference/scalerConstructor.html","id":"arguments-2","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"sigma numeric value matrix.","code":""},{"path":"/reference/scalerConstructor.html","id":"returns","dir":"Reference","previous_headings":"","what":"Returns","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"numeric value matrix","code":""},{"path":"/reference/scalerConstructor.html","id":"method-unscale-beta-","dir":"Reference","previous_headings":"","what":"Method unscale_beta()","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"Unscales beta value (vector) estimated linear model using design matrix scaled object. function works first column initialization data.frame outcome variable.","code":""},{"path":"/reference/scalerConstructor.html","id":"usage-3","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"","code":"scalerConstructor$unscale_beta(beta)"},{"path":"/reference/scalerConstructor.html","id":"arguments-3","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"beta numeric vector beta coefficients estimated linear model.","code":""},{"path":"/reference/scalerConstructor.html","id":"returns-1","dir":"Reference","previous_headings":"","what":"Returns","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"numeric vector.","code":""},{"path":"/reference/scalerConstructor.html","id":"method-clone-","dir":"Reference","previous_headings":"","what":"Method clone()","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"objects class cloneable method.","code":""},{"path":"/reference/scalerConstructor.html","id":"usage-4","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"","code":"scalerConstructor$clone(deep = FALSE)"},{"path":"/reference/scalerConstructor.html","id":"arguments-4","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"deep Whether make deep clone.","code":""},{"path":"/reference/set_simul_pars.html","id":null,"dir":"Reference","previous_headings":"","what":"Set simulation parameters of a study group. — set_simul_pars","title":"Set simulation parameters of a study group. — set_simul_pars","text":"function provides input arguments study group needed simulate data simulate_data(). simulate_data() generates data two-arms clinical trial longitudinal continuous outcomes two intercurrent events (ICEs). ICE1 may thought discontinuation study treatment due study drug condition related (SDCR) reasons. ICE2 may thought discontinuation study treatment due uninformative study drop-, .e. due study drug condition related (NSDRC) reasons outcome data ICE2 always missing.","code":""},{"path":"/reference/set_simul_pars.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Set simulation parameters of a study group. — set_simul_pars","text":"","code":"set_simul_pars( mu, sigma, n, prob_ice1 = 0, or_outcome_ice1 = 1, prob_post_ice1_dropout = 0, prob_ice2 = 0, prob_miss = 0 )"},{"path":"/reference/set_simul_pars.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Set simulation parameters of a study group. — set_simul_pars","text":"mu Numeric vector describing mean outcome trajectory visit (including baseline) assuming ICEs. sigma Covariance matrix outcome trajectory assuming ICEs. n Number subjects belonging group. prob_ice1 Numeric vector specifies probability experiencing ICE1 (discontinuation study treatment due SDCR reasons) visit subject observed outcome visit equal mean baseline (mu[1]). single numeric provided, probability applied visit. or_outcome_ice1 Numeric value specifies odds ratio experiencing ICE1 visit corresponding +1 higher value observed outcome visit. prob_post_ice1_dropout Numeric value specifies probability study drop-following ICE1. subject simulated drop-ICE1, outcomes ICE1 set missing. prob_ice2 Numeric specifies additional probability post-baseline visit affected study drop-. Outcome data subject's first simulated visit affected study drop-subsequent visits set missing. generates second intercurrent event ICE2, may thought treatment discontinuation due NSDRC reasons subsequent drop-. subject, ICE1 ICE2 simulated occur, assumed earlier counts. case ICEs simulated occur time, assumed ICE1 counts. means single subject can experience either ICE1 ICE2, . prob_miss Numeric value specifies additional probability given post-baseline observation missing. can used produce \"intermittent\" missing values associated ICE.","code":""},{"path":"/reference/set_simul_pars.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Set simulation parameters of a study group. — set_simul_pars","text":"simul_pars object named list containing simulation parameters.","code":""},{"path":"/reference/set_simul_pars.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Set simulation parameters of a study group. — set_simul_pars","text":"details, please see simulate_data().","code":""},{"path":[]},{"path":"/reference/set_vars.html","id":null,"dir":"Reference","previous_headings":"","what":"Set key variables — set_vars","title":"Set key variables — set_vars","text":"function used define names key variables within data.frame's provided input arguments draws() ancova().","code":""},{"path":"/reference/set_vars.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Set key variables — set_vars","text":"","code":"set_vars( subjid = \"subjid\", visit = \"visit\", outcome = \"outcome\", group = \"group\", covariates = character(0), strata = group, strategy = \"strategy\" )"},{"path":"/reference/set_vars.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Set key variables — set_vars","text":"subjid name \"Subject ID\" variable. length 1 character vector. visit name \"Visit\" variable. length 1 character vector. outcome name \"Outcome\" variable. length 1 character vector. group name \"Group\" variable. length 1 character vector. covariates name covariates used context modeling. See details. strata name stratification variable used context bootstrap sampling. See details. strategy name \"strategy\" variable. length 1 character vector.","code":""},{"path":"/reference/set_vars.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Set key variables — set_vars","text":"draws() ancova() covariates argument can specified indicate variables included imputation analysis models respectively. wish include interaction terms need manually specified .e. covariates = c(\"group*visit\", \"age*sex\"). Please note use () function inhibit interpretation/conversion objects supported. Currently strata used draws() combination method_condmean(type = \"bootstrap\") method_approxbayes() order allow specification stratified bootstrap sampling. default strata set equal value group assumed users want preserve group size samples. See draws() details. Likewise, currently strategy argument used draws() specify name strategy variable within data_ice data.frame. See draws() details.","code":""},{"path":[]},{"path":"/reference/set_vars.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Set key variables — set_vars","text":"","code":"if (FALSE) { # Using CDISC variable names as an example set_vars( subjid = \"usubjid\", visit = \"avisit\", outcome = \"aval\", group = \"arm\", covariates = c(\"bwt\", \"bht\", \"arm * avisit\"), strategy = \"strat\" ) }"},{"path":"/reference/simulate_data.html","id":null,"dir":"Reference","previous_headings":"","what":"Generate data — simulate_data","title":"Generate data — simulate_data","text":"Generate data two-arms clinical trial longitudinal continuous outcome two intercurrent events (ICEs). ICE1 may thought discontinuation study treatment due study drug condition related (SDCR) reasons. ICE2 may thought discontinuation study treatment due uninformative study drop-, .e. due study drug condition related (NSDRC) reasons outcome data ICE2 always missing.","code":""},{"path":"/reference/simulate_data.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Generate data — simulate_data","text":"","code":"simulate_data(pars_c, pars_t, post_ice1_traj, strategies = getStrategies())"},{"path":"/reference/simulate_data.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Generate data — simulate_data","text":"pars_c simul_pars object generated set_simul_pars(). specifies simulation parameters control arm. pars_t simul_pars object generated set_simul_pars(). specifies simulation parameters treatment arm. post_ice1_traj string specifies observed outcomes occurring ICE1 simulated. Must target function included strategies. Possible choices : Missing Random \"MAR\", Jump Reference \"JR\", Copy Reference \"CR\", Copy Increments Reference \"CIR\", Last Mean Carried Forward \"LMCF\". User-defined strategies also added. See getStrategies() details. strategies named list functions. Default equal getStrategies(). See getStrategies() details.","code":""},{"path":"/reference/simulate_data.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Generate data — simulate_data","text":"data.frame containing simulated data. includes following variables: id: Factor variable specifies id subject. visit: Factor variable specifies visit assessment. Visit 0 denotes baseline visit. group: Factor variable specifies treatment group subject belongs . outcome_bl: Numeric variable specifies baseline outcome. outcome_noICE: Numeric variable specifies longitudinal outcome assuming ICEs. ind_ice1: Binary variable takes value 1 corresponding visit affected ICE1 0 otherwise. dropout_ice1: Binary variable takes value 1 corresponding visit affected drop-following ICE1 0 otherwise. ind_ice2: Binary variable takes value 1 corresponding visit affected ICE2. outcome: Numeric variable specifies longitudinal outcome including ICE1, ICE2 intermittent missing values.","code":""},{"path":"/reference/simulate_data.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Generate data — simulate_data","text":"data generation works follows: Generate outcome data visits (including baseline) multivariate normal distribution parameters pars_c$mu pars_c$sigma control arm parameters pars_t$mu pars_t$sigma treatment arm, respectively. Note randomized trial, outcomes distribution baseline treatment groups, .e. one set pars_c$mu[1]=pars_t$mu[1] pars_c$sigma[1,1]=pars_t$sigma[1,1]. Simulate whether ICE1 (study treatment discontinuation due SDCR reasons) occurs visit according parameters pars_c$prob_ice1 pars_c$or_outcome_ice1 control arm pars_t$prob_ice1 pars_t$or_outcome_ice1 treatment arm, respectively. Simulate drop-following ICE1 according pars_c$prob_post_ice1_dropout pars_t$prob_post_ice1_dropout. Simulate additional uninformative study drop-probabilities pars_c$prob_ice2 pars_t$prob_ice2 visit. generates second intercurrent event ICE2, may thought treatment discontinuation due NSDRC reasons subsequent drop-. simulated time drop-subject's first visit affected drop-data visit subsequent visits consequently set missing. subject, ICE1 ICE2 simulated occur, assumed earlier counts. case ICEs simulated occur time, assumed ICE1 counts. means single subject can experience either ICE1 ICE2, . Adjust trajectories ICE1 according given assumption expressed post_ice1_traj argument. Note post-ICE1 outcomes intervention arm can adjusted. Post-ICE1 outcomes control arm adjusted. Simulate additional intermittent missing outcome data per arguments pars_c$prob_miss pars_t$prob_miss. probability ICE visit modeled according following logistic regression model: ~ 1 + (visit == 0) + ... + (visit == n_visits-1) + ((x-alpha)) : n_visits number visits (including baseline). alpha baseline outcome mean. term ((x-alpha)) specifies dependency probability ICE current outcome value. corresponding regression coefficients logistic model defined follows: intercept set 0, coefficients corresponding discontinuation visit subject outcome equal mean baseline set according parameters pars_c$prob_ice1 (pars_t$prob_ice1), regression coefficient associated covariate ((x-alpha)) set log(pars_c$or_outcome_ice1) (log(pars_t$or_outcome_ice1)). Please note baseline outcome missing affected ICEs.","code":""},{"path":"/reference/simulate_dropout.html","id":null,"dir":"Reference","previous_headings":"","what":"Simulate drop-out — simulate_dropout","title":"Simulate drop-out — simulate_dropout","text":"Simulate drop-","code":""},{"path":"/reference/simulate_dropout.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Simulate drop-out — simulate_dropout","text":"","code":"simulate_dropout(prob_dropout, ids, subset = rep(1, length(ids)))"},{"path":"/reference/simulate_dropout.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Simulate drop-out — simulate_dropout","text":"prob_dropout Numeric specifies probability post-baseline visit affected study drop-. ids Factor variable specifies id subject. subset Binary variable specifies subset affected drop-. .e. subset binary vector length equal length ids takes value 1 corresponding visit affected drop-0 otherwise.","code":""},{"path":"/reference/simulate_dropout.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Simulate drop-out — simulate_dropout","text":"binary vector length equal length ids takes value 1 corresponding outcome affected study drop-.","code":""},{"path":"/reference/simulate_dropout.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Simulate drop-out — simulate_dropout","text":"subset can used specify outcome values affected drop-. default subset set 1 values except values corresponding baseline outcome, since baseline supposed affected drop-. Even subset specified user, values corresponding baseline outcome still hard-coded 0.","code":""},{"path":"/reference/simulate_ice.html","id":null,"dir":"Reference","previous_headings":"","what":"Simulate intercurrent event — simulate_ice","title":"Simulate intercurrent event — simulate_ice","text":"Simulate intercurrent event","code":""},{"path":"/reference/simulate_ice.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Simulate intercurrent event — simulate_ice","text":"","code":"simulate_ice(outcome, visits, ids, prob_ice, or_outcome_ice, baseline_mean)"},{"path":"/reference/simulate_ice.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Simulate intercurrent event — simulate_ice","text":"outcome Numeric variable specifies longitudinal outcome single group. visits Factor variable specifies visit assessment. ids Factor variable specifies id subject. prob_ice Numeric vector specifies visit probability experiencing ICE current visit subject outcome equal mean baseline. single numeric provided, probability applied visit. or_outcome_ice Numeric value specifies odds ratio ICE corresponding +1 higher value outcome visit. baseline_mean Mean outcome value baseline.","code":""},{"path":"/reference/simulate_ice.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Simulate intercurrent event — simulate_ice","text":"binary variable takes value 1 corresponding outcome affected ICE 0 otherwise.","code":""},{"path":"/reference/simulate_ice.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Simulate intercurrent event — simulate_ice","text":"probability ICE visit modeled according following logistic regression model: ~ 1 + (visit == 0) + ... + (visit == n_visits-1) + ((x-alpha)) : n_visits number visits (including baseline). alpha baseline outcome mean set via argument baseline_mean. term ((x-alpha)) specifies dependency probability ICE current outcome value. corresponding regression coefficients logistic model defined follows: intercept set 0, coefficients corresponding discontinuation visit subject outcome equal mean baseline set according parameter or_outcome_ice, regression coefficient associated covariate ((x-alpha)) set log(or_outcome_ice).","code":""},{"path":"/reference/simulate_test_data.html","id":null,"dir":"Reference","previous_headings":"","what":"Create simulated datasets — simulate_test_data","title":"Create simulated datasets — simulate_test_data","text":"Creates longitudinal dataset format rbmi designed analyse.","code":""},{"path":"/reference/simulate_test_data.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create simulated datasets — simulate_test_data","text":"","code":"simulate_test_data( n = 200, sd = c(3, 5, 7), cor = c(0.1, 0.7, 0.4), mu = list(int = 10, age = 3, sex = 2, trt = c(0, 4, 8), visit = c(0, 1, 2)) ) as_vcov(sd, cor)"},{"path":"/reference/simulate_test_data.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create simulated datasets — simulate_test_data","text":"n number subjects sample. Total number observations returned thus n * length(sd) sd standard deviations outcome visit. .e. square root diagonal covariance matrix outcome cor correlation coefficients outcome values visit. See details. mu coefficients use construct mean outcome value visit. Must named list elements int, age, sex, trt & visit. See details.","code":""},{"path":"/reference/simulate_test_data.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Create simulated datasets — simulate_test_data","text":"number visits determined size variance covariance matrix. .e. 3 standard deviation values provided 3 visits per patient created. covariates simulated dataset produced follows: Patients age sampled random N(0,1) distribution Patients sex sampled random 50/50 split Patients group sampled random fixed group n/2 patients outcome variable sampled multivariate normal distribution, see details mean outcome variable derived : coefficients intercept, age sex taken mu$int, mu$age mu$sex respectively, must length 1 numeric. Treatment visit coefficients taken mu$trt mu$visit respectively must either length 1 (.e. constant affect across visits) equal number visits (determined length sd). .e. wanted treatment slope 5 visit slope 1 specify: correlation matrix constructed cor follows. Let cor = c(, b, c, d, e, f) correlation matrix :","code":"outcome = Intercept + age + sex + visit + treatment mu = list(..., \"trt\" = c(0,5,10), \"visit\" = c(0,1,2)) 1 a b d a 1 c e b c 1 f d e f 1"},{"path":"/reference/sort_by.html","id":null,"dir":"Reference","previous_headings":"","what":"Sort data.frame — sort_by","title":"Sort data.frame — sort_by","text":"Sorts data.frame (ascending default) based upon variables within dataset","code":""},{"path":"/reference/sort_by.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Sort data.frame — sort_by","text":"","code":"sort_by(df, vars = NULL, decreasing = FALSE)"},{"path":"/reference/sort_by.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Sort data.frame — sort_by","text":"df data.frame vars character vector variables decreasing logical whether sort order descending ascending (default) order. Can either single logical value (case applied variables) vector length vars","code":""},{"path":"/reference/sort_by.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Sort data.frame — sort_by","text":"","code":"if (FALSE) { sort_by(iris, c(\"Sepal.Length\", \"Sepal.Width\"), decreasing = c(TRUE, FALSE)) }"},{"path":"/reference/split_dim.html","id":null,"dir":"Reference","previous_headings":"","what":"Transform array into list of arrays — split_dim","title":"Transform array into list of arrays — split_dim","text":"Transform array list arrays listing performed given dimension.","code":""},{"path":"/reference/split_dim.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Transform array into list of arrays — split_dim","text":"","code":"split_dim(a, n)"},{"path":"/reference/split_dim.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Transform array into list of arrays — split_dim","text":"Array number dimensions least 2. n Positive integer. Dimension listed.","code":""},{"path":"/reference/split_dim.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Transform array into list of arrays — split_dim","text":"list length n arrays number dimensions equal number dimensions minus 1.","code":""},{"path":"/reference/split_dim.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Transform array into list of arrays — split_dim","text":"example, 3 dimensional array n = 1, split_dim(,n) returns list 2 dimensional arrays (.e. list matrices) element list [, , ], takes values 1 length first dimension array. Example: inputs: <- array( c(1,2,3,4,5,6,7,8,9,10,11,12), dim = c(3,2,2)), means : n <- 1 output res <- split_dim(,n) list 3 elements:","code":"a[1,,] a[2,,] a[3,,] [,1] [,2] [,1] [,2] [,1] [,2] --------- --------- --------- 1 7 2 8 3 9 4 10 5 11 6 12 res[[1]] res[[2]] res[[3]] [,1] [,2] [,1] [,2] [,1] [,2] --------- --------- --------- 1 7 2 8 3 9 4 10 5 11 6 12"},{"path":"/reference/split_imputations.html","id":null,"dir":"Reference","previous_headings":"","what":"Split a flat list of imputation_single() into multiple imputation_df()'s by ID — split_imputations","title":"Split a flat list of imputation_single() into multiple imputation_df()'s by ID — split_imputations","text":"Split flat list imputation_single() multiple imputation_df()'s ID","code":""},{"path":"/reference/split_imputations.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Split a flat list of imputation_single() into multiple imputation_df()'s by ID — split_imputations","text":"","code":"split_imputations(list_of_singles, split_ids)"},{"path":"/reference/split_imputations.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Split a flat list of imputation_single() into multiple imputation_df()'s by ID — split_imputations","text":"list_of_singles list imputation_single()'s split_ids list 1 element per required split. element must contain vector \"ID\"'s correspond imputation_single() ID's required within sample. total number ID's must equal length list_of_singles","code":""},{"path":"/reference/split_imputations.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Split a flat list of imputation_single() into multiple imputation_df()'s by ID — split_imputations","text":"function converts list imputations structured per patient structured per sample .e. converts :","code":"obj <- list( imputation_single(\"Ben\", numeric(0)), imputation_single(\"Ben\", numeric(0)), imputation_single(\"Ben\", numeric(0)), imputation_single(\"Harry\", c(1, 2)), imputation_single(\"Phil\", c(3, 4)), imputation_single(\"Phil\", c(5, 6)), imputation_single(\"Tom\", c(7, 8, 9)) ) index <- list( c(\"Ben\", \"Harry\", \"Phil\", \"Tom\"), c(\"Ben\", \"Ben\", \"Phil\") ) output <- list( imputation_df( imputation_single(id = \"Ben\", values = numeric(0)), imputation_single(id = \"Harry\", values = c(1, 2)), imputation_single(id = \"Phil\", values = c(3, 4)), imputation_single(id = \"Tom\", values = c(7, 8, 9)) ), imputation_df( imputation_single(id = \"Ben\", values = numeric(0)), imputation_single(id = \"Ben\", values = numeric(0)), imputation_single(id = \"Phil\", values = c(5, 6)) ) )"},{"path":"/reference/str_contains.html","id":null,"dir":"Reference","previous_headings":"","what":"Does a string contain a substring — str_contains","title":"Does a string contain a substring — str_contains","text":"Returns vector TRUE/FALSE element x contains element subs .e.","code":"str_contains( c(\"ben\", \"tom\", \"harry\"), c(\"e\", \"y\")) [1] TRUE FALSE TRUE"},{"path":"/reference/str_contains.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Does a string contain a substring — str_contains","text":"","code":"str_contains(x, subs)"},{"path":"/reference/str_contains.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Does a string contain a substring — str_contains","text":"x character vector subs character vector substrings look ","code":""},{"path":"/reference/strategies.html","id":null,"dir":"Reference","previous_headings":"","what":"Strategies — strategies","title":"Strategies — strategies","text":"functions used implement various reference based imputation strategies combining subjects distribution reference distribution based upon visits failed meet Missing--Random (MAR) assumption.","code":""},{"path":"/reference/strategies.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Strategies — strategies","text":"","code":"strategy_MAR(pars_group, pars_ref, index_mar) strategy_JR(pars_group, pars_ref, index_mar) strategy_CR(pars_group, pars_ref, index_mar) strategy_CIR(pars_group, pars_ref, index_mar) strategy_LMCF(pars_group, pars_ref, index_mar)"},{"path":"/reference/strategies.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Strategies — strategies","text":"pars_group list parameters subject's group. See details. pars_ref list parameters subject's reference group. See details. index_mar logical vector indicating visits meet MAR assumption subject. .e. identifies observations non-MAR intercurrent event (ICE).","code":""},{"path":"/reference/strategies.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Strategies — strategies","text":"pars_group pars_ref must list containing elements mu sigma. mu must numeric vector sigma must square matrix symmetric covariance matrix dimensions equal length mu index_mar. e.g. Users can define strategy functions include via strategies argument impute() using getStrategies(). said following strategies available \"box\": Missing Random (MAR) Jump Reference (JR) Copy Reference (CR) Copy Increments Reference (CIR) Last Mean Carried Forward (LMCF)","code":"list( mu = c(1,2,3), sigma = matrix(c(4,3,2,3,5,4,2,4,6), nrow = 3, ncol = 3) )"},{"path":"/reference/string_pad.html","id":null,"dir":"Reference","previous_headings":"","what":"string_pad — string_pad","title":"string_pad — string_pad","text":"Utility function used replicate str_pad. Adds white space either end string get equal desired length","code":""},{"path":"/reference/string_pad.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"string_pad — string_pad","text":"","code":"string_pad(x, width)"},{"path":"/reference/string_pad.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"string_pad — string_pad","text":"x string width desired length","code":""},{"path":"/reference/transpose_imputations.html","id":null,"dir":"Reference","previous_headings":"","what":"Transpose imputations — transpose_imputations","title":"Transpose imputations — transpose_imputations","text":"Takes imputation_df object transposes e.g.","code":"list( list(id = \"a\", values = c(1,2,3)), list(id = \"b\", values = c(4,5,6) ) )"},{"path":"/reference/transpose_imputations.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Transpose imputations — transpose_imputations","text":"","code":"transpose_imputations(imputations)"},{"path":"/reference/transpose_imputations.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Transpose imputations — transpose_imputations","text":"imputations imputation_df object created imputation_df()","code":""},{"path":"/reference/transpose_imputations.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Transpose imputations — transpose_imputations","text":"becomes","code":"list( ids = c(\"a\", \"b\"), values = c(1,2,3,4,5,6) )"},{"path":"/reference/transpose_results.html","id":null,"dir":"Reference","previous_headings":"","what":"Transpose results object — transpose_results","title":"Transpose results object — transpose_results","text":"Transposes Results object (created analyse()) order group estimates together vectors.","code":""},{"path":"/reference/transpose_results.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Transpose results object — transpose_results","text":"","code":"transpose_results(results, components)"},{"path":"/reference/transpose_results.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Transpose results object — transpose_results","text":"results list results. components character vector components extract (.e. \"est\", \"se\").","code":""},{"path":"/reference/transpose_results.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Transpose results object — transpose_results","text":"Essentially function takes object format: produces:","code":"x <- list( list( \"trt1\" = list( est = 1, se = 2 ), \"trt2\" = list( est = 3, se = 4 ) ), list( \"trt1\" = list( est = 5, se = 6 ), \"trt2\" = list( est = 7, se = 8 ) ) ) list( trt1 = list( est = c(1,5), se = c(2,6) ), trt2 = list( est = c(3,7), se = c(4,8) ) )"},{"path":"/reference/transpose_samples.html","id":null,"dir":"Reference","previous_headings":"","what":"Transpose samples — transpose_samples","title":"Transpose samples — transpose_samples","text":"Transposes samples generated draws() grouped subjid instead sample number.","code":""},{"path":"/reference/transpose_samples.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Transpose samples — transpose_samples","text":"","code":"transpose_samples(samples)"},{"path":"/reference/transpose_samples.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Transpose samples — transpose_samples","text":"samples list samples generated draws().","code":""},{"path":"/reference/validate.analysis.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate analysis objects — validate.analysis","title":"Validate analysis objects — validate.analysis","text":"Validates return object analyse() function.","code":""},{"path":"/reference/validate.analysis.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate analysis objects — validate.analysis","text":"","code":"# S3 method for analysis validate(x, ...)"},{"path":"/reference/validate.analysis.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate analysis objects — validate.analysis","text":"x analysis results object (class \"jackknife\", \"bootstrap\", \"rubin\"). ... used.","code":""},{"path":"/reference/validate.draws.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate draws object — validate.draws","title":"Validate draws object — validate.draws","text":"Validate draws object","code":""},{"path":"/reference/validate.draws.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate draws object — validate.draws","text":"","code":"# S3 method for draws validate(x, ...)"},{"path":"/reference/validate.draws.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate draws object — validate.draws","text":"x draws object generated as_draws(). ... used.","code":""},{"path":"/reference/validate.html","id":null,"dir":"Reference","previous_headings":"","what":"Generic validation method — validate","title":"Generic validation method — validate","text":"function used perform assertions object conforms expected structure basic assumptions violated. throw error checks pass.","code":""},{"path":"/reference/validate.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Generic validation method — validate","text":"","code":"validate(x, ...)"},{"path":"/reference/validate.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Generic validation method — validate","text":"x object validated. ... additional arguments pass specific validation method.","code":""},{"path":"/reference/validate.is_mar.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate is_mar for a given subject — validate.is_mar","title":"Validate is_mar for a given subject — validate.is_mar","text":"Checks longitudinal data patient divided MAR followed non-MAR data; non-MAR observation followed MAR observation allowed.","code":""},{"path":"/reference/validate.is_mar.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate is_mar for a given subject — validate.is_mar","text":"","code":"# S3 method for is_mar validate(x, ...)"},{"path":"/reference/validate.is_mar.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate is_mar for a given subject — validate.is_mar","text":"x Object class is_mar. Logical vector indicating whether observations MAR. ... used.","code":""},{"path":"/reference/validate.is_mar.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Validate is_mar for a given subject — validate.is_mar","text":"error issue otherwise return TRUE.","code":""},{"path":"/reference/validate.ivars.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate inputs for vars — validate.ivars","title":"Validate inputs for vars — validate.ivars","text":"Checks required variable names defined within vars appropriate datatypes","code":""},{"path":"/reference/validate.ivars.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate inputs for vars — validate.ivars","text":"","code":"# S3 method for ivars validate(x, ...)"},{"path":"/reference/validate.ivars.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate inputs for vars — validate.ivars","text":"x named list indicating names key variables source dataset ... used","code":""},{"path":"/reference/validate.references.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate user supplied references — validate.references","title":"Validate user supplied references — validate.references","text":"Checks ensure user specified references expect values (.e. found within source data).","code":""},{"path":"/reference/validate.references.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate user supplied references — validate.references","text":"","code":"# S3 method for references validate(x, control, ...)"},{"path":"/reference/validate.references.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate user supplied references — validate.references","text":"x named character vector. control factor variable (group variable source dataset). ... used.","code":""},{"path":"/reference/validate.references.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Validate user supplied references — validate.references","text":"error issue otherwise return TRUE.","code":""},{"path":"/reference/validate.sample_list.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate sample_list object — validate.sample_list","title":"Validate sample_list object — validate.sample_list","text":"Validate sample_list object","code":""},{"path":"/reference/validate.sample_list.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate sample_list object — validate.sample_list","text":"","code":"# S3 method for sample_list validate(x, ...)"},{"path":"/reference/validate.sample_list.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate sample_list object — validate.sample_list","text":"x sample_list object generated sample_list(). ... used.","code":""},{"path":"/reference/validate.sample_single.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate sample_single object — validate.sample_single","title":"Validate sample_single object — validate.sample_single","text":"Validate sample_single object","code":""},{"path":"/reference/validate.sample_single.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate sample_single object — validate.sample_single","text":"","code":"# S3 method for sample_single validate(x, ...)"},{"path":"/reference/validate.sample_single.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate sample_single object — validate.sample_single","text":"x sample_single object generated sample_single(). ... used.","code":""},{"path":"/reference/validate.simul_pars.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate a simul_pars object — validate.simul_pars","title":"Validate a simul_pars object — validate.simul_pars","text":"Validate simul_pars object","code":""},{"path":"/reference/validate.simul_pars.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate a simul_pars object — validate.simul_pars","text":"","code":"# S3 method for simul_pars validate(x, ...)"},{"path":"/reference/validate.simul_pars.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate a simul_pars object — validate.simul_pars","text":"x simul_pars object generated set_simul_pars(). ... used.","code":""},{"path":"/reference/validate.stan_data.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate a stan_data object — validate.stan_data","title":"Validate a stan_data object — validate.stan_data","text":"Validate stan_data object","code":""},{"path":"/reference/validate.stan_data.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate a stan_data object — validate.stan_data","text":"","code":"# S3 method for stan_data validate(x, ...)"},{"path":"/reference/validate.stan_data.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate a stan_data object — validate.stan_data","text":"x stan_data object. ... used.","code":""},{"path":"/reference/validate_analyse_pars.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate analysis results — validate_analyse_pars","title":"Validate analysis results — validate_analyse_pars","text":"Validates analysis results generated analyse().","code":""},{"path":"/reference/validate_analyse_pars.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate analysis results — validate_analyse_pars","text":"","code":"validate_analyse_pars(results, pars)"},{"path":"/reference/validate_analyse_pars.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate analysis results — validate_analyse_pars","text":"results list results generated analysis fun used analyse(). pars list expected parameters analysis. lists .e. c(\"est\", \"se\", \"df\").","code":""},{"path":"/reference/validate_datalong.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate a longdata object — validate_datalong","title":"Validate a longdata object — validate_datalong","text":"Validate longdata object","code":""},{"path":"/reference/validate_datalong.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate a longdata object — validate_datalong","text":"","code":"validate_datalong(data, vars) validate_datalong_varExists(data, vars) validate_datalong_types(data, vars) validate_datalong_notMissing(data, vars) validate_datalong_complete(data, vars) validate_datalong_unifromStrata(data, vars) validate_dataice(data, data_ice, vars, update = FALSE)"},{"path":"/reference/validate_datalong.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate a longdata object — validate_datalong","text":"data data.frame containing longitudinal outcome data + covariates multiple subjects vars vars object created set_vars() data_ice data.frame containing subjects ICE data. See draws() details. update logical, indicates ICE data set first time update applied","code":""},{"path":"/reference/validate_datalong.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Validate a longdata object — validate_datalong","text":"functions used validate various different parts longdata object used draws(), impute(), analyse() pool(). particular: validate_datalong_varExists - Checks variable listed vars actually exists data validate_datalong_types - Checks types key variable expected .e. visit factor variable validate_datalong_notMissing - Checks none key variables (except outcome variable) contain missing values validate_datalong_complete - Checks data complete .e. 1 row subject * visit combination. e.g. nrow(data) == length(unique(subjects)) * length(unique(visits)) validate_datalong_unifromStrata - Checks make sure variables listed stratification variables vary time. e.g. subjects switch stratification groups.","code":""},{"path":"/reference/validate_strategies.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate user specified strategies — validate_strategies","title":"Validate user specified strategies — validate_strategies","text":"Compares user provided strategies required (reference). throw error values reference defined.","code":""},{"path":"/reference/validate_strategies.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate user specified strategies — validate_strategies","text":"","code":"validate_strategies(strategies, reference)"},{"path":"/reference/validate_strategies.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate user specified strategies — validate_strategies","text":"strategies named list strategies. reference list character vector strategies need defined.","code":""},{"path":"/reference/validate_strategies.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Validate user specified strategies — validate_strategies","text":"throw error issue otherwise return TRUE.","code":""},{"path":"/news/index.html","id":"rbmi-development-version","dir":"Changelog","previous_headings":"","what":"rbmi (development version)","title":"rbmi (development version)","text":"Include vignette obtain frequentist information-anchored inference conditional mean imputation using rbmi Added FAQ vignette","code":""},{"path":"/news/index.html","id":"rbmi-126","dir":"Changelog","previous_headings":"","what":"rbmi 1.2.6","title":"rbmi 1.2.6","text":"CRAN release: 2023-11-24 Updated unit tests fix false-positive error CRAN’s testing servers","code":""},{"path":"/news/index.html","id":"rbmi-125","dir":"Changelog","previous_headings":"","what":"rbmi 1.2.5","title":"rbmi 1.2.5","text":"CRAN release: 2023-09-20 Updated internal Stan code ensure future compatibility (@andrjohns, #390) Updated package description include relevant references (#393) Fixed documentation typos (#393)","code":""},{"path":"/news/index.html","id":"rbmi-123","dir":"Changelog","previous_headings":"","what":"rbmi 1.2.3","title":"rbmi 1.2.3","text":"CRAN release: 2022-11-14 Minor internal tweaks ensure compatibility packages rbmi depends ","code":""},{"path":"/news/index.html","id":"rbmi-121","dir":"Changelog","previous_headings":"","what":"rbmi 1.2.1","title":"rbmi 1.2.1","text":"CRAN release: 2022-10-25 Removed native pipes |> testing code package backwards compatible older servers Replaced glmmTMB dependency mmrm package. resulted package stable (less model fitting convergence issues) well speeding run times 3-fold.","code":""},{"path":"/news/index.html","id":"rbmi-114","dir":"Changelog","previous_headings":"","what":"rbmi 1.1.4","title":"rbmi 1.1.4","text":"CRAN release: 2022-05-18 Updated urls references vignettes Fixed bug visit factor levels re-constructed incorrectly delta_template() Fixed bug wrong visit displayed error message specific visit doesn’t data draws() Fixed bug wrong input parameter displayed error message simulate_data()","code":""},{"path":"/news/index.html","id":"rbmi-111--113","dir":"Changelog","previous_headings":"","what":"rbmi 1.1.1 & 1.1.3","title":"rbmi 1.1.1 & 1.1.3","text":"CRAN release: 2022-03-08 change functionality 1.1.0 Various minor tweaks address CRAN checks messages","code":""},{"path":"/news/index.html","id":"rbmi-110","dir":"Changelog","previous_headings":"","what":"rbmi 1.1.0","title":"rbmi 1.1.0","text":"CRAN release: 2022-03-02 Initial public release","code":""}] +[{"path":"/CONTRIBUTING.html","id":null,"dir":"","previous_headings":"","what":"Contributing to rbmi","title":"Contributing to rbmi","text":"file outlines propose make changes rbmi well providing details obscure aspects package’s development process.","code":""},{"path":"/CONTRIBUTING.html","id":"setup","dir":"","previous_headings":"","what":"Setup","title":"Contributing to rbmi","text":"order develop contribute rbmi need access C/C++ compiler. Windows install rtools macOS install Xcode. Likewise, also need install package’s development dependencies. can done launching R within project root executing:","code":"devtools::install_dev_deps()"},{"path":"/CONTRIBUTING.html","id":"code-changes","dir":"","previous_headings":"","what":"Code changes","title":"Contributing to rbmi","text":"want make code contribution, ’s good idea first file issue make sure someone team agrees ’s needed. ’ve found bug, please file issue illustrates bug minimal reprex (also help write unit test, needed).","code":""},{"path":"/CONTRIBUTING.html","id":"pull-request-process","dir":"","previous_headings":"Code changes","what":"Pull request process","title":"Contributing to rbmi","text":"project uses simple GitHub flow model development. , code changes done feature branch based main branch merged back main branch complete. Pull Requests accepted unless CI/CD checks passed. (See CI/CD section information). Pull Requests relating package’s core R code must accompanied corresponding unit test. pull requests containing changes core R code contain unit test demonstrate working intended accepted. (See Unit Testing section information). Pull Requests add lines changed NEWS.md file.","code":""},{"path":"/CONTRIBUTING.html","id":"coding-considerations","dir":"","previous_headings":"Code changes","what":"Coding Considerations","title":"Contributing to rbmi","text":"use roxygen2, Markdown syntax, documentation. Please ensure code conforms lintr. can check running lintr::lint(\"FILE NAME\") files modified ensuring findings kept possible. hard requirements following lintr’s conventions encourage developers follow guidance closely possible. project uses 4 space indents, contributions following accepted. project makes use S3 R6 OOP. Usage S4 OOP systems avoided unless absolutely necessary ensure consistency. said recommended stick S3 unless modification place R6 specific features required. current desire package keep dependency tree small possible. end discouraged adding additional packages “Depends” / “Imports” section unless absolutely essential. importing package just use single function consider just copying source code function instead, though please check licence include proper attribution/notices. expectations “Suggests” free use package vignettes / unit tests, though please mindful unnecessarily excessive .","code":""},{"path":"/CONTRIBUTING.html","id":"unit-testing--cicd","dir":"","previous_headings":"","what":"Unit Testing & CI/CD","title":"Contributing to rbmi","text":"project uses testthat perform unit testing combination GitHub Actions CI/CD.","code":""},{"path":"/CONTRIBUTING.html","id":"scheduled-testing","dir":"","previous_headings":"Unit Testing & CI/CD","what":"Scheduled Testing","title":"Contributing to rbmi","text":"Due stochastic nature package unit tests take considerable amount time execute. avoid issues usability, unit tests take couple seconds run deferred scheduled testing. tests run occasionally periodic basis (currently twice month) every pull request / push event. defer test scheduled build simply include skip_if_not(is_full_test()) top test_that() block .e. scheduled tests can also manually activated going “https://github.com/insightsengineering/rbmi” -> “Actions” -> “Bi-Weekly” -> “Run Workflow”. advisable releasing CRAN.","code":"test_that(\"some unit test\", { skip_if_not(is_full_test()) expect_equal(1,1) })"},{"path":"/CONTRIBUTING.html","id":"cran-releases","dir":"","previous_headings":"Unit Testing & CI/CD","what":"CRAN Releases","title":"Contributing to rbmi","text":"order release package CRAN needs tested across multiple different OS’s versions R. implemented project via GitHub Action Workflow titled “Check CRAN” needs manually activated. go “https://github.com/insightsengineering/rbmi” -> “Actions” -> “Check CRAN” -> “Run Workflow”. tests pass package can safely released CRAN (updating relevant cran-comments.md file)","code":""},{"path":"/CONTRIBUTING.html","id":"docker-images","dir":"","previous_headings":"Unit Testing & CI/CD","what":"Docker Images","title":"Contributing to rbmi","text":"support CI/CD terms reducing installation time, several Docker images pre-built contain packages system dependencies project needs. current relevant images can found : ghcr.io/insightsengineering/rbmi:r404 ghcr.io/insightsengineering/rbmi:r410 ghcr.io/insightsengineering/rbmi:latest latest image automatically re-built month contain latest version R packages. versions built older versions R (indicated tag number) contain package versions version R released. important ensure package works older versions R many companies typically run due delays validation processes. code create images can found misc/docker. legacy images (.e. everything excluding “latest” image) built manual request running corresponding GitHub Actions Workflow.","code":""},{"path":"/CONTRIBUTING.html","id":"reproducibility-print-tests--snaps","dir":"","previous_headings":"Unit Testing & CI/CD","what":"Reproducibility, Print Tests & Snaps","title":"Contributing to rbmi","text":"particular issue testing package reproducibility. part handled well via set.seed() however stan/rstan guarantee reproducibility even seed run different hardware. issue surfaces testing print messages pool object displays treatment estimates thus identical run different machines. address issue pre-made pool objects generated stored R/sysdata.rda (generated data-raw/create_print_test_data.R). generated print messages compared expected values stored tests/testthat/_snaps/ (automatically created testthat::expect_snapshot())","code":""},{"path":"/CONTRIBUTING.html","id":"fitting-mmrms","dir":"","previous_headings":"","what":"Fitting MMRM’s","title":"Contributing to rbmi","text":"package currently uses mmrm package fit MMRM models. package still fairly new far proven stable, fast reliable. spot issues MMRM package please raise corresponding GitHub Repository - link mmrm package uses TMB uncommon see warnings either inconsistent versions TMB Matrix package compiled . order resolve may wish re-compile packages source using: Note need rtools installed Windows machine Xcode running macOS (somehow else access C/C++ compiler).","code":"install.packages(c(\"TMB\", \"mmrm\"), type = \"source\")"},{"path":"/CONTRIBUTING.html","id":"rstan","dir":"","previous_headings":"","what":"rstan","title":"Contributing to rbmi","text":"Bayesian models fitted package implemented via stan/rstan. code can found inst/stan/MMRM.stan. Note package automatically take care compiling code install run devtools::load_all(). Please note package won’t recompile code unless changed source code delete src directory.","code":""},{"path":"/CONTRIBUTING.html","id":"vignettes","dir":"","previous_headings":"","what":"Vignettes","title":"Contributing to rbmi","text":"CRAN imposes 10-minute run limit building, compiling testing package. keep limit vignettes pre-built; say simply changing source code automatically update vignettes, need manually re-build . need run: re-built need commit updated *.html files git repository. reference static vignette process works using “asis” vignette engine provided R.rsp. works getting R recognise vignettes files ending *.html.asis; builds simply copying corresponding files ending *.html relevent docs/ folder built package.","code":"Rscript vignettes/build.R"},{"path":"/CONTRIBUTING.html","id":"misc--local-folders","dir":"","previous_headings":"","what":"Misc & Local Folders","title":"Contributing to rbmi","text":"misc/ folder project used hold useful scripts, analyses, simulations & infrastructure code wish keep isn’t essential build deployment package. Feel free store additional stuff feel worth keeping. Likewise, local/ added .gitignore file meaning anything stored folder won’t committed repository. example, may find useful storing personal scripts testing generally exploring package development.","code":""},{"path":"/LICENSE.html","id":null,"dir":"","previous_headings":"","what":"Apache License","title":"Apache License","text":"Version 2.0, January 2004 ","code":""},{"path":[]},{"path":"/LICENSE.html","id":"id_1-definitions","dir":"","previous_headings":"Terms and Conditions for use, reproduction, and distribution","what":"1. Definitions","title":"Apache License","text":"“License” shall mean terms conditions use, reproduction, distribution defined Sections 1 9 document. “Licensor” shall mean copyright owner entity authorized copyright owner granting License. “Legal Entity” shall mean union acting entity entities control, controlled , common control entity. purposes definition, “control” means () power, direct indirect, cause direction management entity, whether contract otherwise, (ii) ownership fifty percent (50%) outstanding shares, (iii) beneficial ownership entity. “” (“”) shall mean individual Legal Entity exercising permissions granted License. “Source” form shall mean preferred form making modifications, including limited software source code, documentation source, configuration files. “Object” form shall mean form resulting mechanical transformation translation Source form, including limited compiled object code, generated documentation, conversions media types. “Work” shall mean work authorship, whether Source Object form, made available License, indicated copyright notice included attached work (example provided Appendix ). “Derivative Works” shall mean work, whether Source Object form, based (derived ) Work editorial revisions, annotations, elaborations, modifications represent, whole, original work authorship. purposes License, Derivative Works shall include works remain separable , merely link (bind name) interfaces , Work Derivative Works thereof. “Contribution” shall mean work authorship, including original version Work modifications additions Work Derivative Works thereof, intentionally submitted Licensor inclusion Work copyright owner individual Legal Entity authorized submit behalf copyright owner. purposes definition, “submitted” means form electronic, verbal, written communication sent Licensor representatives, including limited communication electronic mailing lists, source code control systems, issue tracking systems managed , behalf , Licensor purpose discussing improving Work, excluding communication conspicuously marked otherwise designated writing copyright owner “Contribution.” “Contributor” shall mean Licensor individual Legal Entity behalf Contribution received Licensor subsequently incorporated within Work.","code":""},{"path":"/LICENSE.html","id":"id_2-grant-of-copyright-license","dir":"","previous_headings":"Terms and Conditions for use, reproduction, and distribution","what":"2. Grant of Copyright License","title":"Apache License","text":"Subject terms conditions License, Contributor hereby grants perpetual, worldwide, non-exclusive, -charge, royalty-free, irrevocable copyright license reproduce, prepare Derivative Works , publicly display, publicly perform, sublicense, distribute Work Derivative Works Source Object form.","code":""},{"path":"/LICENSE.html","id":"id_3-grant-of-patent-license","dir":"","previous_headings":"Terms and Conditions for use, reproduction, and distribution","what":"3. Grant of Patent License","title":"Apache License","text":"Subject terms conditions License, Contributor hereby grants perpetual, worldwide, non-exclusive, -charge, royalty-free, irrevocable (except stated section) patent license make, made, use, offer sell, sell, import, otherwise transfer Work, license applies patent claims licensable Contributor necessarily infringed Contribution(s) alone combination Contribution(s) Work Contribution(s) submitted. institute patent litigation entity (including cross-claim counterclaim lawsuit) alleging Work Contribution incorporated within Work constitutes direct contributory patent infringement, patent licenses granted License Work shall terminate date litigation filed.","code":""},{"path":"/LICENSE.html","id":"id_4-redistribution","dir":"","previous_headings":"Terms and Conditions for use, reproduction, and distribution","what":"4. Redistribution","title":"Apache License","text":"may reproduce distribute copies Work Derivative Works thereof medium, without modifications, Source Object form, provided meet following conditions: () must give recipients Work Derivative Works copy License; (b) must cause modified files carry prominent notices stating changed files; (c) must retain, Source form Derivative Works distribute, copyright, patent, trademark, attribution notices Source form Work, excluding notices pertain part Derivative Works; (d) Work includes “NOTICE” text file part distribution, Derivative Works distribute must include readable copy attribution notices contained within NOTICE file, excluding notices pertain part Derivative Works, least one following places: within NOTICE text file distributed part Derivative Works; within Source form documentation, provided along Derivative Works; , within display generated Derivative Works, wherever third-party notices normally appear. contents NOTICE file informational purposes modify License. may add attribution notices within Derivative Works distribute, alongside addendum NOTICE text Work, provided additional attribution notices construed modifying License. may add copyright statement modifications may provide additional different license terms conditions use, reproduction, distribution modifications, Derivative Works whole, provided use, reproduction, distribution Work otherwise complies conditions stated License.","code":""},{"path":"/LICENSE.html","id":"id_5-submission-of-contributions","dir":"","previous_headings":"Terms and Conditions for use, reproduction, and distribution","what":"5. Submission of Contributions","title":"Apache License","text":"Unless explicitly state otherwise, Contribution intentionally submitted inclusion Work Licensor shall terms conditions License, without additional terms conditions. Notwithstanding , nothing herein shall supersede modify terms separate license agreement may executed Licensor regarding Contributions.","code":""},{"path":"/LICENSE.html","id":"id_6-trademarks","dir":"","previous_headings":"Terms and Conditions for use, reproduction, and distribution","what":"6. Trademarks","title":"Apache License","text":"License grant permission use trade names, trademarks, service marks, product names Licensor, except required reasonable customary use describing origin Work reproducing content NOTICE file.","code":""},{"path":"/LICENSE.html","id":"id_7-disclaimer-of-warranty","dir":"","previous_headings":"Terms and Conditions for use, reproduction, and distribution","what":"7. Disclaimer of Warranty","title":"Apache License","text":"Unless required applicable law agreed writing, Licensor provides Work (Contributor provides Contributions) “” BASIS, WITHOUT WARRANTIES CONDITIONS KIND, either express implied, including, without limitation, warranties conditions TITLE, NON-INFRINGEMENT, MERCHANTABILITY, FITNESS PARTICULAR PURPOSE. solely responsible determining appropriateness using redistributing Work assume risks associated exercise permissions License.","code":""},{"path":"/LICENSE.html","id":"id_8-limitation-of-liability","dir":"","previous_headings":"Terms and Conditions for use, reproduction, and distribution","what":"8. Limitation of Liability","title":"Apache License","text":"event legal theory, whether tort (including negligence), contract, otherwise, unless required applicable law (deliberate grossly negligent acts) agreed writing, shall Contributor liable damages, including direct, indirect, special, incidental, consequential damages character arising result License use inability use Work (including limited damages loss goodwill, work stoppage, computer failure malfunction, commercial damages losses), even Contributor advised possibility damages.","code":""},{"path":"/LICENSE.html","id":"id_9-accepting-warranty-or-additional-liability","dir":"","previous_headings":"Terms and Conditions for use, reproduction, and distribution","what":"9. Accepting Warranty or Additional Liability","title":"Apache License","text":"redistributing Work Derivative Works thereof, may choose offer, charge fee , acceptance support, warranty, indemnity, liability obligations /rights consistent License. However, accepting obligations, may act behalf sole responsibility, behalf Contributor, agree indemnify, defend, hold Contributor harmless liability incurred , claims asserted , Contributor reason accepting warranty additional liability. END TERMS CONDITIONS","code":""},{"path":"/LICENSE.html","id":"appendix-how-to-apply-the-apache-license-to-your-work","dir":"","previous_headings":"","what":"APPENDIX: How to apply the Apache License to your work","title":"Apache License","text":"apply Apache License work, attach following boilerplate notice, fields enclosed brackets [] replaced identifying information. (Don’t include brackets!) text enclosed appropriate comment syntax file format. also recommend file class name description purpose included “printed page” copyright notice easier identification within third-party archives.","code":"Copyright [yyyy] [name of copyright owner] Licensed under the Apache License, Version 2.0 (the \"License\"); you may not use this file except in compliance with the License. You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0 Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an \"AS IS\" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License."},{"path":"/articles/CondMean_Inference.html","id":"introduction","dir":"Articles","previous_headings":"","what":"Introduction","title":"rbmi: Inference with Conditional Mean Imputation","text":"described section 3.10.2 statistical specifications package (vignette(topic = \"stat_specs\", package = \"rbmi\")), two different types variance estimators proposed reference-based imputation methods statistical literature (Bartlett (2023)). first frequentist variance describes actual repeated sampling variability estimator results inference correct frequentist sense, .e. hypothesis tests accurate type error control confidence intervals correct coverage probabilities repeated sampling reference-based assumption correctly specified (Bartlett (2023), Wolbers et al. (2022)). Reference-based missing data assumption strong borrow information control arm imputation active arm. consequence, size frequentist standard errors treatment effects may decrease increasing amounts missing data. second -called “information-anchored” variance originally proposed context sensitivity analyses (Cro, Carpenter, Kenward (2019)). variance estimator based disentangling point estimation variance estimation altogether. resulting information-anchored variance typically similar variance missing--random (MAR) imputation increases increasing amounts missing data approximately rate MAR imputation. However, information-anchored variance reflect actual variability reference-based estimator resulting frequentist inference highly conservative resulting substantial power loss. Reference-based conditional mean imputation combined resampling method jackknife bootstrap first introduced Wolbers et al. (2022). approach naturally targets frequentist variance. information-anchored variance typically estimated using Rubin’s rules Bayesian multiple imputation applicable within conditional mean imputation framework. However, alternative information-anchored variance proposed Lu (2021) can easily obtained show . basic idea Lu (2021) obtain information-anchored variance via MAR imputation combined delta-adjustment delta selected data-driven way match reference-based estimator. conditional mean imputation, proposal Lu (2021) can implemented choosing delta-adjustment difference conditional mean imputation chosen reference-based assumption MAR original dataset. variance can obtained via jackknife bootstrap keeping delta-adjustment fixed. resulting variance estimate similar Rubin’s variance. Moreover, shown Cro, Carpenter, Kenward (2019), variance MAR-imputation combined delta-adjustment achieves even better information-anchoring properties Rubin’s variance reference-based imputation. Reference-based missing data assumptions strong borrow information control arm imputation active arm. vignette demonstrates first obtain frequentist inference using reference-based conditional mean imputation using rbmi, shows information-anchored inference can also easily implemented using package.","code":""},{"path":"/articles/CondMean_Inference.html","id":"data-and-model-specification","dir":"Articles","previous_headings":"","what":"Data and model specification","title":"rbmi: Inference with Conditional Mean Imputation","text":"use publicly available example dataset antidepressant clinical trial active drug versus placebo. relevant endpoint Hamilton 17-item depression rating scale (HAMD17) assessed baseline weeks 1, 2, 4, 6. Study drug discontinuation occurred 24% subjects active drug 26% subjects placebo. data study drug discontinuation missing single additional intermittent missing observation. consider imputation model mean change baseline HAMD17 score outcome (variable CHANGE dataset). following covariates included imputation model: treatment group (THERAPY), (categorical) visit (VISIT), treatment--visit interactions, baseline HAMD17 score (BASVAL), baseline HAMD17 score--visit interactions. common unstructured covariance matrix structure assumed groups. analysis model ANCOVA model treatment group primary factor adjustment baseline HAMD17 score. example, assume imputation strategy ICE “study-drug discontinuation” Jump Reference (JR) subjects imputation based conditional mean imputation combined jackknife resampling (bootstrap also selected).","code":""},{"path":"/articles/CondMean_Inference.html","id":"reference-based-conditional-mean-imputation---frequentist-inference","dir":"Articles","previous_headings":"","what":"Reference-based conditional mean imputation - frequentist inference","title":"rbmi: Inference with Conditional Mean Imputation","text":"Conditional mean imputation combined resampling method jackknife bootstrap naturally targets frequentist estimation standard error treatment effect, thus providing valid frequentist inference. provide code obtain frequentist inference reference-based conditional mean imputation using rbmi. code used section almost identical code quickstart vignette (vignette(topic = \"quickstart\", package = \"rbmi\")) except use conditional mean imputation combined jackknife (method_condmean(type = \"jackknife\")) rather Bayesian multiple imputation (method_bayes()). therefore refer vignette help files individual functions explanations details.","code":""},{"path":"/articles/CondMean_Inference.html","id":"draws","dir":"Articles","previous_headings":"3 Reference-based conditional mean imputation - frequentist inference","what":"Draws","title":"rbmi: Inference with Conditional Mean Imputation","text":"make use rbmi::expand_locf() expand dataset order one row per subject per visit missing outcomes denoted NA. construct data_ice, vars method input arguments first core rbmi function, draws(). Finally, call function draws() derive parameter estimates base imputation model full dataset leave-one-subject-samples.","code":"library(rbmi) library(dplyr) #> #> Attaching package: 'dplyr' #> The following objects are masked from 'package:stats': #> #> filter, lag #> The following objects are masked from 'package:base': #> #> intersect, setdiff, setequal, union dat <- antidepressant_data # Use expand_locf to add rows corresponding to visits with missing outcomes to # the dataset dat <- expand_locf( dat, PATIENT = levels(dat$PATIENT), # expand by PATIENT and VISIT VISIT = levels(dat$VISIT), vars = c(\"BASVAL\", \"THERAPY\"), # fill with LOCF BASVAL and THERAPY group = c(\"PATIENT\"), order = c(\"PATIENT\", \"VISIT\") ) # create data_ice and set the imputation strategy to JR for # each patient with at least one missing observation dat_ice <- dat %>% arrange(PATIENT, VISIT) %>% filter(is.na(CHANGE)) %>% group_by(PATIENT) %>% slice(1) %>% ungroup() %>% select(PATIENT, VISIT) %>% mutate(strategy = \"JR\") # In this dataset, subject 3618 has an intermittent missing values which # does not correspond to a study drug discontinuation. We therefore remove # this subject from `dat_ice`. (In the later imputation step, it will # automatically be imputed under the default MAR assumption.) dat_ice <- dat_ice[-which(dat_ice$PATIENT == 3618),] # Define the names of key variables in our dataset and # the covariates included in the imputation model using `set_vars()` vars <- set_vars( outcome = \"CHANGE\", visit = \"VISIT\", subjid = \"PATIENT\", group = \"THERAPY\", covariates = c(\"BASVAL*VISIT\", \"THERAPY*VISIT\") ) # Define which imputation method to use (here: conditional mean imputation # with jackknife as resampling) method <- method_condmean(type = \"jackknife\") # Create samples for the imputation parameters by running the draws() function drawObj <- draws( data = dat, data_ice = dat_ice, vars = vars, method = method, quiet = TRUE ) drawObj #> #> Draws Object #> ------------ #> Number of Samples: 1 + 172 #> Number of Failed Samples: 0 #> Model Formula: CHANGE ~ 1 + THERAPY + VISIT + BASVAL * VISIT + THERAPY * VISIT #> Imputation Type: condmean #> Method: #> name: Conditional Mean #> covariance: us #> threshold: 0.01 #> same_cov: TRUE #> REML: TRUE #> type: jackknife"},{"path":"/articles/CondMean_Inference.html","id":"impute","dir":"Articles","previous_headings":"3 Reference-based conditional mean imputation - frequentist inference","what":"Impute","title":"rbmi: Inference with Conditional Mean Imputation","text":"can use now function impute() perform imputation original dataset leave-one-samples using results obtained previous step.","code":"references <- c(\"DRUG\" = \"PLACEBO\", \"PLACEBO\" = \"PLACEBO\") imputeObj <- impute(drawObj, references) imputeObj #> #> Imputation Object #> ----------------- #> Number of Imputed Datasets: 1 + 172 #> Fraction of Missing Data (Original Dataset): #> 4: 0% #> 5: 8% #> 6: 13% #> 7: 25% #> References: #> DRUG -> PLACEBO #> PLACEBO -> PLACEBO"},{"path":"/articles/CondMean_Inference.html","id":"analyse","dir":"Articles","previous_headings":"3 Reference-based conditional mean imputation - frequentist inference","what":"Analyse","title":"rbmi: Inference with Conditional Mean Imputation","text":"datasets imputed, can call analyse() function apply complete-data analysis model (ANCOVA) imputed dataset.","code":"# Set analysis variables using rbmi function \"set_vars\" vars_an <- set_vars( group = vars$group, visit = vars$visit, outcome = vars$outcome, covariates = \"BASVAL\" ) # Analyse MAR imputation with derived delta adjustment anaObj <- analyse( imputeObj, rbmi::ancova, vars = vars_an ) anaObj #> #> Analysis Object #> --------------- #> Number of Results: 1 + 172 #> Analysis Function: rbmi::ancova #> Delta Applied: FALSE #> Analysis Estimates: #> trt_4 #> lsm_ref_4 #> lsm_alt_4 #> trt_5 #> lsm_ref_5 #> lsm_alt_5 #> trt_6 #> lsm_ref_6 #> lsm_alt_6 #> trt_7 #> lsm_ref_7 #> lsm_alt_7"},{"path":"/articles/CondMean_Inference.html","id":"pool","dir":"Articles","previous_headings":"3 Reference-based conditional mean imputation - frequentist inference","what":"Pool","title":"rbmi: Inference with Conditional Mean Imputation","text":"Finally, can extract treatment effect estimates perform inference using jackknife variance estimator. done calling pool() function. gives estimated treatment effect 2.13 (95% CI 0.44 3.81) last visit associated p-value 0.013.","code":"poolObj <- pool(anaObj) poolObj #> #> Pool Object #> ----------- #> Number of Results Combined: 1 + 172 #> Method: jackknife #> Confidence Level: 0.95 #> Alternative: two.sided #> #> Results: #> #> ================================================== #> parameter est se lci uci pval #> -------------------------------------------------- #> trt_4 -0.092 0.695 -1.453 1.27 0.895 #> lsm_ref_4 -1.616 0.588 -2.767 -0.464 0.006 #> lsm_alt_4 -1.708 0.396 -2.484 -0.931 <0.001 #> trt_5 1.305 0.878 -0.416 3.027 0.137 #> lsm_ref_5 -4.133 0.688 -5.481 -2.785 <0.001 #> lsm_alt_5 -2.828 0.604 -4.011 -1.645 <0.001 #> trt_6 1.929 0.862 0.239 3.619 0.025 #> lsm_ref_6 -6.088 0.671 -7.402 -4.773 <0.001 #> lsm_alt_6 -4.159 0.686 -5.503 -2.815 <0.001 #> trt_7 2.126 0.858 0.444 3.807 0.013 #> lsm_ref_7 -6.965 0.685 -8.307 -5.622 <0.001 #> lsm_alt_7 -4.839 0.762 -6.333 -3.346 <0.001 #> --------------------------------------------------"},{"path":"/articles/CondMean_Inference.html","id":"reference-based-conditional-mean-imputation---information-anchored-inference","dir":"Articles","previous_headings":"","what":"Reference-based conditional mean imputation - information-anchored inference","title":"rbmi: Inference with Conditional Mean Imputation","text":"section, present estimation process based conditional mean imputation combined jackknife can adapted obtain information-anchored variance following proposal Lu (2021).","code":""},{"path":"/articles/CondMean_Inference.html","id":"draws-1","dir":"Articles","previous_headings":"4 Reference-based conditional mean imputation - information-anchored inference","what":"Draws","title":"rbmi: Inference with Conditional Mean Imputation","text":"code pre-processing dataset “draws” step equivalent code provided frequentist inference. Please refer section details step.","code":"library(rbmi) library(dplyr) dat <- antidepressant_data # Use expand_locf to add rows corresponding to visits with missing outcomes to # the dataset dat <- expand_locf( dat, PATIENT = levels(dat$PATIENT), # expand by PATIENT and VISIT VISIT = levels(dat$VISIT), vars = c(\"BASVAL\", \"THERAPY\"), # fill with LOCF BASVAL and THERAPY group = c(\"PATIENT\"), order = c(\"PATIENT\", \"VISIT\") ) # create data_ice and set the imputation strategy to JR for # each patient with at least one missing observation dat_ice <- dat %>% arrange(PATIENT, VISIT) %>% filter(is.na(CHANGE)) %>% group_by(PATIENT) %>% slice(1) %>% ungroup() %>% select(PATIENT, VISIT) %>% mutate(strategy = \"JR\") # In this dataset, subject 3618 has an intermittent missing values which # does not correspond to a study drug discontinuation. We therefore remove # this subject from `dat_ice`. (In the later imputation step, it will # automatically be imputed under the default MAR assumption.) dat_ice <- dat_ice[-which(dat_ice$PATIENT == 3618),] # Define the names of key variables in our dataset and # the covariates included in the imputation model using `set_vars()` vars <- set_vars( outcome = \"CHANGE\", visit = \"VISIT\", subjid = \"PATIENT\", group = \"THERAPY\", covariates = c(\"BASVAL*VISIT\", \"THERAPY*VISIT\") ) # Define which imputation method to use (here: conditional mean imputation # with jackknife as resampling) method <- method_condmean(type = \"jackknife\") # Create samples for the imputation parameters by running the draws() function drawObj <- draws( data = dat, data_ice = dat_ice, vars = vars, method = method, quiet = TRUE ) drawObj"},{"path":"/articles/CondMean_Inference.html","id":"imputation-step-including-calculation-of-delta-adjustment","dir":"Articles","previous_headings":"4 Reference-based conditional mean imputation - information-anchored inference","what":"Imputation step including calculation of delta-adjustment","title":"rbmi: Inference with Conditional Mean Imputation","text":"proposal Lu (2021) replace reference-based imputation MAR imputation combined delta-adjustment delta selected data-driven way match reference-based estimator. rbmi, implemented first performing imputation defined reference-based imputation strategy (JR) well MAR separately. Second, delta-adjustment defined difference conditional mean imputation reference-based MAR imputation, respectively, original dataset. simplify implementation, written function get_delta_match_refBased performs step. function takes input arguments draws object, data_ice (.e. data.frame containing information intercurrent events imputation strategies), references, named vector identifies references used reference-based imputation methods. function returns list containing imputation objects reference-based MAR imputation, plus data.frame contains delta-adjustment.","code":"#' Get delta adjustment that matches reference-based imputation #' #' @param draws: A `draws` object created by `draws()`. #' @param data_ice: `data.frame` containing the information about the intercurrent #' events and the imputation strategies. Must represent the desired imputation #' strategy and not the MAR-variant. #' @param references: A named vector. Identifies the references to be used #' for reference-based imputation methods. #' #' @return #' The function returns a list containing the imputation objects under both #' reference-based and MAR imputation, plus a `data.frame` which contains the #' delta-adjustment. #' #' @seealso `draws()`, `impute()`. get_delta_match_refBased <- function(draws, data_ice, references) { # Impute according to `data_ice` imputeObj <- impute( draws = drawObj, update_strategy = data_ice, references = references ) vars <- imputeObj$data$vars # Access imputed dataset (index=1 for method_condmean(type = \"jackknife\")) cmi <- extract_imputed_dfs(imputeObj, index = 1, idmap = TRUE)[[1]] idmap <- attributes(cmi)$idmap cmi <- cmi[, c(vars$subjid, vars$visit, vars$outcome)] colnames(cmi)[colnames(cmi) == vars$outcome] <- \"y_imp\" # Map back original patients id since rbmi re-code ids to ensure id uniqueness cmi[[vars$subjid]] <- idmap[match(cmi[[vars$subjid]], names(idmap))] # Derive conditional mean imputations under MAR dat_ice_MAR <- data_ice dat_ice_MAR[[vars$strategy]] <- \"MAR\" # Impute under MAR # Note that in this specific context, it is desirable that an update # from a reference-based strategy to MAR uses the exact same data for # fitting the imputation models, i.e. that available post-ICE data are # omitted from the imputation model for both. This is the case when # using argument update_strategy in function impute(). # However, for other settings (i.e. if one is interested in switching to # a standard MAR imputation strategy altogether), this behavior is # undesirable and, consequently, the function throws a warning which # we suppress here. suppressWarnings( imputeObj_MAR <- impute( draws, update_strategy = dat_ice_MAR ) ) # Access imputed dataset (index=1 for method_condmean(type = \"jackknife\")) cmi_MAR <- extract_imputed_dfs(imputeObj_MAR, index = 1, idmap = TRUE)[[1]] idmap <- attributes(cmi_MAR)$idmap cmi_MAR <- cmi_MAR[, c(vars$subjid, vars$visit, vars$outcome)] colnames(cmi_MAR)[colnames(cmi_MAR) == vars$outcome] <- \"y_MAR\" # Map back original patients id since rbmi re-code ids to ensure id uniqueness cmi_MAR[[vars$subjid]] <- idmap[match(cmi_MAR[[vars$subjid]], names(idmap))] # Derive delta adjustment \"aligned with ref-based imputation\", # i.e. difference between ref-based imputation and MAR imputation delta_adjust <- merge(cmi, cmi_MAR, by = c(vars$subjid, vars$visit), all = TRUE) delta_adjust$delta <- delta_adjust$y_imp - delta_adjust$y_MAR ret_obj <- list( imputeObj = imputeObj, imputeObj_MAR = imputeObj_MAR, delta_adjust = delta_adjust ) return(ret_obj) } references <- c(\"DRUG\" = \"PLACEBO\", \"PLACEBO\" = \"PLACEBO\") res_delta_adjust <- get_delta_match_refBased(drawObj, dat_ice, references)"},{"path":"/articles/CondMean_Inference.html","id":"analyse-1","dir":"Articles","previous_headings":"4 Reference-based conditional mean imputation - information-anchored inference","what":"Analyse","title":"rbmi: Inference with Conditional Mean Imputation","text":"use function analyse() add delta-adjustment perform analysis imputed datasets MAR. analyse() take input argument imputations = res_delta_adjust$imputeObj_MAR, .e. imputation object corresponding MAR imputation (JR imputation). argument delta can used add delta-adjustment prior analysis set delta-adjustment obtained previous step: delta = res_delta_adjust$delta_adjust.","code":"# Set analysis variables using rbmi function \"set_vars\" vars_an <- set_vars( group = vars$group, visit = vars$visit, outcome = vars$outcome, covariates = \"BASVAL\" ) # Analyse MAR imputation with derived delta adjustment anaObj_MAR_delta <- analyse( res_delta_adjust$imputeObj_MAR, rbmi::ancova, delta = res_delta_adjust$delta_adjust, vars = vars_an )"},{"path":"/articles/CondMean_Inference.html","id":"pool-1","dir":"Articles","previous_headings":"4 Reference-based conditional mean imputation - information-anchored inference","what":"Pool","title":"rbmi: Inference with Conditional Mean Imputation","text":"can finally use pool() function extract treatment effect estimate (well estimated marginal means) visit apply jackknife variance estimator analysis estimates imputed leave-one-samples. gives estimated treatment effect 2.13 (95% CI -0.08 4.33) last visit associated p-value 0.058. Per construction delta-adjustment, point estimate identical frequentist analysis. However, standard error much larger (1.12 vs. 0.86). Indeed, information-anchored standard error (resulting inference) similar results Baysesian multiple imputation using Rubin’s rules standard error 1.13 reported quickstart vignette (vignette(topic = \"quickstart\", package = \"rbmi\"). note, shown e.g. Wolbers et al. (2022), hypothesis testing based information-anchored inference conservative, .e. actual type error much lower nominal value. Hence, confidence intervals \\(p\\)-values based information-anchored inference interpreted caution.","code":"poolObj_MAR_delta <- pool(anaObj_MAR_delta) poolObj_MAR_delta #> #> Pool Object #> ----------- #> Number of Results Combined: 1 + 172 #> Method: jackknife #> Confidence Level: 0.95 #> Alternative: two.sided #> #> Results: #> #> ================================================== #> parameter est se lci uci pval #> -------------------------------------------------- #> trt_4 -0.092 0.695 -1.453 1.27 0.895 #> lsm_ref_4 -1.616 0.588 -2.767 -0.464 0.006 #> lsm_alt_4 -1.708 0.396 -2.484 -0.931 <0.001 #> trt_5 1.305 0.944 -0.545 3.156 0.167 #> lsm_ref_5 -4.133 0.738 -5.579 -2.687 <0.001 #> lsm_alt_5 -2.828 0.603 -4.01 -1.646 <0.001 #> trt_6 1.929 0.993 -0.018 3.876 0.052 #> lsm_ref_6 -6.088 0.758 -7.574 -4.602 <0.001 #> lsm_alt_6 -4.159 0.686 -5.504 -2.813 <0.001 #> trt_7 2.126 1.123 -0.076 4.327 0.058 #> lsm_ref_7 -6.965 0.85 -8.63 -5.299 <0.001 #> lsm_alt_7 -4.839 0.763 -6.335 -3.343 <0.001 #> --------------------------------------------------"},{"path":[]},{"path":"/articles/FAQ.html","id":"introduction","dir":"Articles","previous_headings":"","what":"Introduction","title":"rbmi: Frequently Asked Questions","text":"document provides answers common questions rbmi package. intended read rbmi: Quickstart vignette.","code":""},{"path":"/articles/FAQ.html","id":"how-do-the-methods-in-rbmi-compare-to-the-mixed-model-for-repeated-measures-mmrm-implemented-in-the-mmrm-package","dir":"Articles","previous_headings":"1 Introduction","what":"How do the methods in rbmi compare to the mixed model for repeated measures (MMRM) implemented in the mmrm package?","title":"rbmi: Frequently Asked Questions","text":"rbmi designed complement , occasionally, replace standard MMRM analyses clinical trials longitudinal endpoints. Strengths rbmi compared standard MMRM model : rbmi designed allow analyses fully aligned estimand definition. facilitate , implements methods range different missing data assumptions including standard missing--random (MAR), extended MAR (via inclusion time-varying covariates), reference-based missingness, missing--random random (NMAR; via \\(\\delta\\)-adjustments). contrast, standard MMRM model valid standard MAR assumption always plausible. example, standard MAR assumption rather implausible implementing treatment policy strategy intercurrent event “treatment discontinuation” substantial proportion subjects lost--follow-discontinuation. \\(\\delta\\)-adjustment methods implemented rbmi can used sensitivity analyses primary MMRM- rbmi-type analysis. Weaknesses rbmi compared standard MMRM model : MMRM models de-facto standard analysis method decade. rbmi currently less established. rbmi computationally intensive using requires careful planning.","code":""},{"path":"/articles/FAQ.html","id":"how-does-rbmi-compare-to-general-purpose-software-for-multiple-imputation-mi-such-as-mice","dir":"Articles","previous_headings":"1 Introduction","what":"How does rbmi compare to general-purpose software for multiple imputation (MI) such as mice?","title":"rbmi: Frequently Asked Questions","text":"rbmi covers “MMRM-type” settings, .e. settings single longitudinal continuous outcome may missing visits hence require imputation. settings, several advantages general-purpose MI software: rbmi supports imputation range different missing data assumptions whereas general-purpose MI software mostly focused MAR-based imputation. particular, unclear implement jump reference (JR) copy increments reference (CIR) methods software. rbmi interface fully streamlined setting arguably makes implementation straightforward general-purpose MI software. MICE algorithm stochastic inference always based Rubin’s rules. contrast, method “conditional mean imputation plus jackknifing” (method=\"method_condmean(type = \"jackknife\")\") rbmi require tuning parameters, fully deterministic, provides frequentist-consistent inference also reference-based imputations (Rubin’s rule conservative leading actual type error rates can far nominal values). However, rbmi much limited functionality general-purpose MI software.","code":""},{"path":"/articles/FAQ.html","id":"how-to-handle-missing-data-in-baseline-covariates-in-rbmi","dir":"Articles","previous_headings":"1 Introduction","what":"How to handle missing data in baseline covariates in rbmi?","title":"rbmi: Frequently Asked Questions","text":"rbmi support imputation missing baseline covariates. Therefore, missing baseline covariates need handled outside rbmi. best approach handling missing baseline covariates needs made case--case basis context randomized trials, relatively simple approach often sufficient (White Thompson (2005)).","code":""},{"path":"/articles/FAQ.html","id":"why-does-rbmi-by-default-use-an-ancova-analysis-model-and-not-an-mmrm-analysis-model","dir":"Articles","previous_headings":"1 Introduction","what":"Why does rbmi by default use an ANCOVA analysis model and not an MMRM analysis model?","title":"rbmi: Frequently Asked Questions","text":"theoretical justification conditional mean imputation method requires analysis model leads point estimator linear function outcome vector (Wolbers et al. (2022)). case ANCOVA general MMRM models. imputation methods, ANCOVA MMRM valid analysis methods. MMRM analysis model implemented providing custom analysis function analyse() function. expalanations, also cite end section 2.4 conditional mean imputation paper (Wolbers et al. (2022)): proof relies fact ANCOVA estimator linear function outcome vector. complete data, ANCOVA estimator leads identical parameter estimates MMRM model longitudinal outcomes arbitrary common covariance structure across treatment groups treatment--visit interactions well covariate--visit-interactions included analysis model covariates,17 (p. 197). Hence, proof also applies MMRM models. expect conditional mean imputation also valid general MMRM model used analysis involved argument required formally justify .","code":""},{"path":"/articles/FAQ.html","id":"how-can-i-analyse-the-change-from-baseline-in-the-analysis-model-when-imputation-was-done-on-the-original-outcomes","dir":"Articles","previous_headings":"1 Introduction","what":"How can I analyse the change-from-baseline in the analysis model when imputation was done on the original outcomes?","title":"rbmi: Frequently Asked Questions","text":"can achieved using custom analysis functions outlined Section 7 Advanced Vignette. e.g.","code":"ancova_modified <- function(data, ...) { data2 <- data %>% mutate(ENDPOINT = ENDPOINT - BASELINE) rbmi::ancova(data2, ...) } anaObj <- rbmi::analyse( imputeObj, ancova_modified, vars = vars )"},{"path":"/articles/advanced.html","id":"introduction","dir":"Articles","previous_headings":"","what":"Introduction","title":"rbmi: Advanced Functionality","text":"purpose vignette provide overview advanced features rbmi package. sections vignette relatively self-contained, .e. readers able jump directly section covers functionality interested .","code":""},{"path":"/articles/advanced.html","id":"sec:dataSimul","dir":"Articles","previous_headings":"","what":"Data simulation using function simulate_data()","title":"rbmi: Advanced Functionality","text":"order demonstrate advanced functions first create simulated dataset rbmi function simulate_data(). simulate_data() function generates data randomized clinical trial longitudinal continuous outcomes two different types intercurrent events (ICEs). One intercurrent event (ICE1) may thought discontinuation study treatment due study drug condition related (SDCR) reasons. event (ICE2) may thought discontinuation study treatment due study drug condition related (NSDCR) reasons. purpose vignette, simulate data similarly simulation study reported Wolbers et al. (2022) (though change simulation parameters) include one ICE type (ICE1). Specifically, simulate 1:1 randomized trial active drug (intervention) versus placebo (control) 100 subjects per group 6 post-baseline assessments (bi-monthly visits 12 months) following assumptions: mean outcome trajectory placebo group increases linearly 50 baseline (visit 0) 60 visit 6, .e. slope 10 points/year. mean outcome trajectory intervention group identical placebo group visit 2. visit 2 onward, slope decreases 50% 5 points/year. covariance structure baseline follow-values groups implied random intercept slope model standard deviation 5 intercept slope, correlation 0.25. addition, independent residual error standard deviation 2.5 added assessment. probability study drug discontinuation visit calculated according logistic model depends observed outcome visit. Specifically, visit-wise discontinuation probability 2% 3% control intervention group, respectively, specified case observed outcome equal 50 (mean value baseline). odds discontinuation simulated increase +10% +1 point increase observed outcome. Study drug discontinuation simulated effect mean trajectory placebo group. intervention group, subjects discontinue follow slope mean trajectory placebo group time point onward. compatible copy increments reference (CIR) assumption. Study drop-study drug discontinuation visit occurs probability 50% leading missing outcome data time point onward. function simulate_data() requires 3 arguments (see function documentation help(simulate_data) details): pars_c: simulation parameters control group pars_t: simulation parameters intervention group post_ice1_traj: Specifies observed outcomes ICE1 simulated , report data according specifications can simulated function simulate_data():","code":"library(rbmi) library(dplyr) library(ggplot2) library(purrr) set.seed(122) n <- 100 time <- c(0, 2, 4, 6, 8, 10, 12) # Mean trajectory control muC <- c(50.0, 51.66667, 53.33333, 55.0, 56.66667, 58.33333, 60.0) # Mean trajectory intervention muT <- c(50.0, 51.66667, 53.33333, 54.16667, 55.0, 55.83333, 56.66667) # Create Sigma sd_error <- 2.5 covRE <- rbind( c(25.0, 6.25), c(6.25, 25.0) ) Sigma <- cbind(1, time / 12) %*% covRE %*% rbind(1, time / 12) + diag(sd_error^2, nrow = length(time)) # Set probability of discontinuation probDisc_C <- 0.02 probDisc_T <- 0.03 or_outcome <- 1.10 # +1 point increase => +10% odds of discontinuation # Set drop-out rate following discontinuation prob_dropout <- 0.5 # Set simulation parameters of the control group parsC <- set_simul_pars( mu = muC, sigma = Sigma, n = n, prob_ice1 = probDisc_C, or_outcome_ice1 = or_outcome, prob_post_ice1_dropout = prob_dropout ) # Set simulation parameters of the intervention group parsT <- parsC parsT$mu <- muT parsT$prob_ice1 <- probDisc_T # Set assumption about post-ice trajectory post_ice_traj <- \"CIR\" # Simulate data data <- simulate_data( pars_c = parsC, pars_t = parsT, post_ice1_traj = post_ice_traj ) head(data) #> id visit group outcome_bl outcome_noICE ind_ice1 ind_ice2 dropout_ice1 #> 1 id_1 0 Control 57.32704 57.32704 0 0 0 #> 2 id_1 1 Control 57.32704 54.69751 1 0 1 #> 3 id_1 2 Control 57.32704 58.60702 1 0 1 #> 4 id_1 3 Control 57.32704 61.50119 1 0 1 #> 5 id_1 4 Control 57.32704 56.68363 1 0 1 #> 6 id_1 5 Control 57.32704 66.14799 1 0 1 #> outcome #> 1 57.32704 #> 2 NA #> 3 NA #> 4 NA #> 5 NA #> 6 NA # As a simple descriptive of the simulated data, summarize the number of subjects with ICEs and missing data data %>% group_by(id) %>% summarise( group = group[1], any_ICE = (any(ind_ice1 == 1)), any_NA = any(is.na(outcome))) %>% group_by(group) %>% summarise( subjects_with_ICE = sum(any_ICE), subjects_with_missings = sum(any_NA) ) #> # A tibble: 2 × 3 #> group subjects_with_ICE subjects_with_missings #> #> 1 Control 18 8 #> 2 Intervention 25 14"},{"path":"/articles/advanced.html","id":"sec:postICEobs","dir":"Articles","previous_headings":"","what":"Handling of observed post-ICE data in rbmi under reference-based imputation","title":"rbmi: Advanced Functionality","text":"rbmi always uses non-missing outcome data input data set, .e. data never overwritten imputation step removed analysis step. implies data considered irrelevant treatment effect estimation (e.g. data ICE estimand specified hypothetical strategy), data need removed input data set user prior calling rbmi functions. imputation missing random (MAR) strategy, observed outcome data also included fitting base imputation model. However, ICEs handled using reference-based imputation methods (CIR, CR, JR), rbmi excludes observed post-ICE data base imputation model. data excluded, base imputation model mistakenly estimate mean trajectories based mixture observed pre- post-ICE data relevant reference-based imputations. However, observed post-ICE data added back data set fitting base imputation model included subsequent imputation analysis steps. Post-ICE data control reference group also excluded base imputation model user specifies reference-based imputation strategy ICEs. ensures ICE impact data included base imputation model regardless whether ICE occurred control intervention group. hand, imputation reference group based MAR assumption even reference-based imputation methods may preferable settings include post-ICE data control group base imputation model. can implemented specifying MAR strategy ICE control group reference-based strategy ICE intervention group. use latter approach example . simulated trial data section 2 assumed outcomes intervention group observed ICE “treatment discontinuation” follow increments observed control group. Thus imputation missing data intervention group treatment discontinuation might performed reference-based copy increments reference (CIR) assumption. Specifically, implement estimator following assumptions: endpoint interest change outcome baseline visit. imputation model includes treatment group, (categorical) visit, treatment--visit interactions, baseline outcome, baseline outcome--visit interactions covariates. imputation model assumes common unstructured covariance matrix treatment groups control group, missing data imputed MAR whereas intervention group, missing post-ICE data imputed CIR assumption analysis model endpoint imputed datasets separate ANCOVA model visit treatment group primary covariate adjustment baseline outcome value. illustration purposes, chose MI based approximate Bayesian posterior draws 20 random imputations demanding computational perspective. practical applications, number random imputations may need increased. Moreover, imputations also supported rbmi. guidance regarding choice imputation approach, refer user comparison implemented approaches Section 3.9 “Statistical Specifications” vignette (vignette(\"stat_specs\", package = \"rbmi\")). first report code set variables imputation analysis models. yet familiar syntax, recommend first check “quickstart” vignette (vignette(\"quickstart\", package = \"rbmi\")). chosen imputation method can set function method_approxbayes() follows: can now sequentially call 4 key functions rbmi perform multiple imputation. Please note management observed post-ICE data performed without additional complexity user. draws() automatically excludes post-ICE data handled reference-based method (keeps post-ICE data handled using MAR) using information provided argument data_ice. impute() impute truly missing data data[[vars$outcome]]. last output gives estimated difference -4.537 (95% CI -6.420 -2.655) two groups last visit associated p-value lower 0.001.","code":"# Create data_ice including the subject's first visit affected by the ICE and the imputation strategy # Imputation strategy for post-ICE data is CIR in the intervention group and MAR for the control group # (note that ICEs which are handled using MAR are optional and do not impact the analysis # because imputation of missing data under MAR is the default) data_ice_CIR <- data %>% group_by(id) %>% filter(ind_ice1 == 1) %>% # select visits with ICEs mutate(strategy = ifelse(group == \"Intervention\", \"CIR\", \"MAR\")) %>% summarise( visit = visit[1], # Select first visit affected by the ICE strategy = strategy[1] ) # Compute endpoint of interest: change from baseline and # remove rows corresponding to baseline visits data <- data %>% filter(visit != 0) %>% mutate( change = outcome - outcome_bl, visit = factor(visit, levels = unique(visit)) ) # Define key variables for the imputation and analysis models vars <- set_vars( subjid = \"id\", visit = \"visit\", outcome = \"change\", group = \"group\", covariates = c(\"visit*outcome_bl\", \"visit*group\"), strategy = \"strategy\" ) vars_an <- vars vars_an$covariates <- \"outcome_bl\" method <- method_approxbayes(n_sample = 20) draw_obj <- draws( data = data, data_ice = data_ice_CIR, vars = vars, method = method, quiet = TRUE, ncores = 2 ) impute_obj_CIR <- impute( draw_obj, references = c(\"Control\" = \"Control\", \"Intervention\" = \"Control\") ) ana_obj_CIR <- analyse( impute_obj_CIR, vars = vars_an ) pool_obj_CIR <- pool(ana_obj_CIR) pool_obj_CIR #> #> Pool Object #> ----------- #> Number of Results Combined: 20 #> Method: rubin #> Confidence Level: 0.95 #> Alternative: two.sided #> #> Results: #> #> ================================================== #> parameter est se lci uci pval #> -------------------------------------------------- #> trt_1 -0.486 0.512 -1.496 0.524 0.343 #> lsm_ref_1 2.62 0.362 1.907 3.333 <0.001 #> lsm_alt_1 2.133 0.362 1.42 2.847 <0.001 #> trt_2 -0.066 0.542 -1.135 1.004 0.904 #> lsm_ref_2 3.707 0.384 2.95 4.464 <0.001 #> lsm_alt_2 3.641 0.383 2.885 4.397 <0.001 #> trt_3 -1.782 0.607 -2.979 -0.585 0.004 #> lsm_ref_3 5.841 0.428 4.997 6.685 <0.001 #> lsm_alt_3 4.059 0.428 3.214 4.904 <0.001 #> trt_4 -2.518 0.692 -3.884 -1.152 <0.001 #> lsm_ref_4 7.656 0.492 6.685 8.627 <0.001 #> lsm_alt_4 5.138 0.488 4.176 6.1 <0.001 #> trt_5 -3.658 0.856 -5.346 -1.97 <0.001 #> lsm_ref_5 9.558 0.598 8.379 10.737 <0.001 #> lsm_alt_5 5.9 0.608 4.699 7.101 <0.001 #> trt_6 -4.537 0.954 -6.42 -2.655 <0.001 #> lsm_ref_6 11.048 0.666 9.735 12.362 <0.001 #> lsm_alt_6 6.511 0.674 5.181 7.841 <0.001 #> --------------------------------------------------"},{"path":"/articles/advanced.html","id":"efficiently-changing-reference-based-imputation-strategies","dir":"Articles","previous_headings":"","what":"Efficiently changing reference-based imputation strategies","title":"rbmi: Advanced Functionality","text":"draws() function far computationally intensive function rbmi. settings, may important explore impact change reference-based imputation strategy results. change affect imputation model affect subsequent imputation step. order allow changes imputation strategy without re-run draws() function, function impute() additional argument update_strategies. However, please note functionality comes important limitations: described beginning Section 3, post-ICE outcomes included input dataset base imputation model imputation method MAR excluded reference-based imputation methods (CIR, CR, JR). Therefore, updata_strategies applied imputation strategy changed MAR non-MAR strategy presence observed post-ICE outcomes. Similarly, change non-MAR strategy MAR triggers warning presence observed post-ICE outcomes base imputation model fitted relevant data MAR. Finally, update_strategies applied timing ICEs changed (argument data_ice) addition imputation strategy. example, described analysis copy increments reference (CIR) assumption previous section. Let’s assume want change strategy jump reference imputation strategy sensitivity analysis. can efficiently implemented using update_strategies follows: imputations jump reference assumption, get estimated difference -4.360 (95% CI -6.238 -2.482) two groups last visit associated p-value <0.001.","code":"# Change ICE strategy from CIR to JR data_ice_JR <- data_ice_CIR %>% mutate(strategy = ifelse(strategy == \"CIR\", \"JR\", strategy)) impute_obj_JR <- impute( draw_obj, references = c(\"Control\" = \"Control\", \"Intervention\" = \"Control\"), update_strategy = data_ice_JR ) ana_obj_JR <- analyse( impute_obj_JR, vars = vars_an ) pool_obj_JR <- pool(ana_obj_JR) pool_obj_JR #> #> Pool Object #> ----------- #> Number of Results Combined: 20 #> Method: rubin #> Confidence Level: 0.95 #> Alternative: two.sided #> #> Results: #> #> ================================================== #> parameter est se lci uci pval #> -------------------------------------------------- #> trt_1 -0.485 0.513 -1.496 0.526 0.346 #> lsm_ref_1 2.609 0.363 1.892 3.325 <0.001 #> lsm_alt_1 2.124 0.361 1.412 2.836 <0.001 #> trt_2 -0.06 0.535 -1.115 0.995 0.911 #> lsm_ref_2 3.694 0.378 2.948 4.441 <0.001 #> lsm_alt_2 3.634 0.381 2.882 4.387 <0.001 #> trt_3 -1.767 0.598 -2.948 -0.587 0.004 #> lsm_ref_3 5.845 0.422 5.012 6.677 <0.001 #> lsm_alt_3 4.077 0.432 3.225 4.93 <0.001 #> trt_4 -2.529 0.686 -3.883 -1.175 <0.001 #> lsm_ref_4 7.637 0.495 6.659 8.614 <0.001 #> lsm_alt_4 5.108 0.492 4.138 6.078 <0.001 #> trt_5 -3.523 0.856 -5.212 -1.833 <0.001 #> lsm_ref_5 9.554 0.61 8.351 10.758 <0.001 #> lsm_alt_5 6.032 0.611 4.827 7.237 <0.001 #> trt_6 -4.36 0.952 -6.238 -2.482 <0.001 #> lsm_ref_6 11.003 0.676 9.669 12.337 <0.001 #> lsm_alt_6 6.643 0.687 5.287 8 <0.001 #> --------------------------------------------------"},{"path":"/articles/advanced.html","id":"imputation-under-mar-with-time-varying-covariates","dir":"Articles","previous_headings":"","what":"Imputation under MAR with time-varying covariates","title":"rbmi: Advanced Functionality","text":"Guizzaro et al. (2021) suggested implement treatment policy strategy via imputation MAR assumption conditioning subject’s ICE status, .e. impute missing post-ICE data based observed post-ICE data. One possible implementation proposal add time-varying covariates imputation model. case study implements proposal compares reference-based imputation methods estimators early Parkinson’s disease can found Noci et al. (2021). settings, may carried including binary time-varying indicator subject’s ICE status visit (defined 0 pre-ICE visits 1 post-ICE visits) imputation model. However, simulated data introduced section 2, may plausible assume treatment discontinuation leads change “slope” mean outcome trajectory. can implemented including time-varying covariate equal 0 visits prior treatment discontinuation equal time treatment discontinuation subsequent visits. regression coefficient corresponding change post-ICE “slope” allowed depend assigned treatment group, .e. imputation model include interaction time-varying covariate treatment group. Let’s first define time-varying covariate: can include time-varying covariate imputation model, crossed group variable: now sequentially call 4 key rbmi functions:","code":"data <- data %>% group_by(id) %>% mutate(time_from_ice1 = cumsum(ind_ice1)*2/12 ) # multiplication by 2/12 because visits are bi-monthly vars_tv <- set_vars( subjid = \"id\", visit = \"visit\", outcome = \"change\", group = \"group\", covariates = c(\"visit*outcome_bl\", \"visit*group\", \"time_from_ice1*group\"), strategy = \"strategy\" ) draw_obj <- draws( data = data, data_ice = NULL, # if NULL, MAR is assumed for all missing data vars = vars_tv, method = method, quiet = TRUE ) impute_obj_tv <- impute( draw_obj, references = c(\"Control\" = \"Control\", \"Intervention\" = \"Intervention\") ) ana_obj_tv <- analyse( impute_obj_tv, vars = vars_an ) pool(ana_obj_tv) #> #> Pool Object #> ----------- #> Number of Results Combined: 20 #> Method: rubin #> Confidence Level: 0.95 #> Alternative: two.sided #> #> Results: #> #> ================================================== #> parameter est se lci uci pval #> -------------------------------------------------- #> trt_1 -0.492 0.515 -1.507 0.524 0.341 #> lsm_ref_1 2.623 0.362 1.908 3.338 <0.001 #> lsm_alt_1 2.131 0.366 1.409 2.854 <0.001 #> trt_2 0.018 0.55 -1.067 1.103 0.974 #> lsm_ref_2 3.697 0.382 2.943 4.45 <0.001 #> lsm_alt_2 3.715 0.394 2.936 4.493 <0.001 #> trt_3 -1.802 0.614 -3.015 -0.59 0.004 #> lsm_ref_3 5.815 0.429 4.97 6.661 <0.001 #> lsm_alt_3 4.013 0.441 3.142 4.884 <0.001 #> trt_4 -2.543 0.704 -3.932 -1.154 <0.001 #> lsm_ref_4 7.609 0.486 6.65 8.568 <0.001 #> lsm_alt_4 5.066 0.516 4.046 6.086 <0.001 #> trt_5 -3.739 0.879 -5.475 -2.004 <0.001 #> lsm_ref_5 9.499 0.606 8.302 10.695 <0.001 #> lsm_alt_5 5.759 0.636 4.502 7.017 <0.001 #> trt_6 -4.685 0.98 -6.622 -2.748 <0.001 #> lsm_ref_6 10.988 0.667 9.67 12.305 <0.001 #> lsm_alt_6 6.302 0.712 4.894 7.711 <0.001 #> --------------------------------------------------"},{"path":"/articles/advanced.html","id":"custom-imputation-strategies","dir":"Articles","previous_headings":"","what":"Custom imputation strategies","title":"rbmi: Advanced Functionality","text":"following imputation strategies implemented rbmi: Missing Random (MAR) Jump Reference (JR) Copy Reference (CR) Copy Increments Reference (CIR) Last Mean Carried Forward (LMCF) addition, rbmi allows user implement imputation strategy. , user needs three things: Define function implementing new imputation strategy. Specify patients use strategy data_ice dataset provided draws(). Provide imputation strategy function impute(). imputation strategy function must take 3 arguments (pars_group, pars_ref, index_mar) calculates mean covariance matrix subject’s marginal imputation distribution applied subjects strategy applies. , pars_group contains predicted mean trajectory (pars_group$mu, numeric vector) covariance matrix (pars_group$sigma) subject conditional assigned treatment group covariates. pars_ref contains corresponding mean trajectory covariance matrix conditional reference group subject’s covariates. index_mar logical vector specifies visit whether visit unaffected ICE handled using non-MAR method . example, user can check CIR strategy implemented looking function strategy_CIR(). illustrate simple example, assume new strategy implemented follows: - marginal mean imputation distribution equal marginal mean trajectory subject according assigned group covariates ICE. - ICE marginal mean imputation distribution equal average visit-wise marginal means based subjects covariates assigned group reference group, respectively. - covariance matrix marginal imputation distribution, covariance matrix assigned group taken. , first need define imputation function example coded follows: example showing use: incorporate rbmi, data_ice needs updated strategy AVG specified visits affected ICE. Additionally, function needs provided impute() via getStrategies() function shown : , analysis proceed calling analyse() pool() .","code":"strategy_CIR #> function (pars_group, pars_ref, index_mar) #> { #> if (all(index_mar)) { #> return(pars_group) #> } #> else if (all(!index_mar)) { #> return(pars_ref) #> } #> mu <- pars_group$mu #> last_mar <- which(!index_mar)[1] - 1 #> increments_from_last_mar_ref <- pars_ref$mu[!index_mar] - #> pars_ref$mu[last_mar] #> mu[!index_mar] <- mu[last_mar] + increments_from_last_mar_ref #> sigma <- compute_sigma(sigma_group = pars_group$sigma, sigma_ref = pars_ref$sigma, #> index_mar = index_mar) #> pars <- list(mu = mu, sigma = sigma) #> return(pars) #> } #> #> strategy_AVG <- function(pars_group, pars_ref, index_mar) { mu_mean <- (pars_group$mu + pars_ref$mu) / 2 x <- pars_group x$mu[!index_mar] <- mu_mean[!index_mar] return(x) } pars_group <- list( mu = c(1, 2, 3), sigma = as_vcov(c(1, 3, 2), c(0.4, 0.5, 0.45)) ) pars_ref <- list( mu = c(5, 6, 7), sigma = as_vcov(c(2, 1, 1), c(0.7, 0.8, 0.5)) ) index_mar <- c(TRUE, TRUE, FALSE) strategy_AVG(pars_group, pars_ref, index_mar) #> $mu #> [1] 1 2 5 #> #> $sigma #> [,1] [,2] [,3] #> [1,] 1.0 1.2 1.0 #> [2,] 1.2 9.0 2.7 #> [3,] 1.0 2.7 4.0 data_ice_AVG <- data_ice_CIR %>% mutate(strategy = ifelse(strategy == \"CIR\", \"AVG\", strategy)) draw_obj <- draws( data = data, data_ice = data_ice_AVG, vars = vars, method = method, quiet = TRUE ) impute_obj <- impute( draw_obj, references = c(\"Control\" = \"Control\", \"Intervention\" = \"Control\"), strategies = getStrategies(AVG = strategy_AVG) )"},{"path":"/articles/advanced.html","id":"custom-analysis-functions","dir":"Articles","previous_headings":"","what":"Custom analysis functions","title":"rbmi: Advanced Functionality","text":"default rbmi analyse data using ancova() function. analysis function fits ANCOVA model outcomes visit separately, returns “treatment effect” estimate well corresponding least square means group. user wants perform different analysis, return different statistics analysis, can done using custom analysis function. Beware validity conditional mean imputation method formally established analysis functions corresponding linear models (ANCOVA) caution required applying alternative analysis functions method. custom analysis function must take data.frame first argument return named list element list containing minimum point estimate, called est. method method_bayes() method_approxbayes(), list must additionally contain standard error (element se) , available, degrees freedom complete-data analysis model (element df). simple example, replicate ANCOVA analysis last visit CIR-based imputations user-defined analysis function : second example, assume supplementary analysis user wants compare proportion subjects change baseline >10 points last visit treatment groups baseline outcome additional covariate. lead following basic analysis function: Note user wants rbmi use normal approximation pooled test statistics, degrees freedom need set df = NA (per example). degrees freedom complete data test statistics known degrees freedom set df = Inf, rbmi pools degrees freedom across imputed datasets according rule Barnard Rubin (see “Statistical Specifications” vignette (vignette(\"stat_specs\", package = \"rbmi\") details). According rule, infinite degrees freedom complete data analysis imply pooled degrees freedom also infinite. Rather, case pooled degrees freedom (M-1)/lambda^2, M number imputations lambda fraction missing information (see Barnard Rubin (1999) details).","code":"compare_change_lastvisit <- function(data, ...) { fit <- lm(change ~ group + outcome_bl, data = data, subset = (visit == 6) ) res <- list( trt = list( est = coef(fit)[\"groupIntervention\"], se = sqrt(vcov(fit)[\"groupIntervention\", \"groupIntervention\"]), df = df.residual(fit) ) ) return(res) } ana_obj_CIR6 <- analyse( impute_obj_CIR, fun = compare_change_lastvisit, vars = vars_an ) pool(ana_obj_CIR6) #> #> Pool Object #> ----------- #> Number of Results Combined: 20 #> Method: rubin #> Confidence Level: 0.95 #> Alternative: two.sided #> #> Results: #> #> ================================================= #> parameter est se lci uci pval #> ------------------------------------------------- #> trt -4.537 0.954 -6.42 -2.655 <0.001 #> ------------------------------------------------- compare_prop_lastvisit <- function(data, ...) { fit <- glm( I(change > 10) ~ group + outcome_bl, family = binomial(), data = data, subset = (visit == 6) ) res <- list( trt = list( est = coef(fit)[\"groupIntervention\"], se = sqrt(vcov(fit)[\"groupIntervention\", \"groupIntervention\"]), df = NA ) ) return(res) } ana_obj_prop <- analyse( impute_obj_CIR, fun = compare_prop_lastvisit, vars = vars_an ) pool_obj_prop <- pool(ana_obj_prop) pool_obj_prop #> #> Pool Object #> ----------- #> Number of Results Combined: 20 #> Method: rubin #> Confidence Level: 0.95 #> Alternative: two.sided #> #> Results: #> #> ================================================= #> parameter est se lci uci pval #> ------------------------------------------------- #> trt -1.052 0.314 -1.667 -0.438 0.001 #> ------------------------------------------------- tmp <- as.data.frame(pool_obj_prop) %>% mutate( OR = exp(est), OR.lci = exp(lci), OR.uci = exp(uci) ) %>% select(parameter, OR, OR.lci, OR.uci) tmp #> parameter OR OR.lci OR.uci #> 1 trt 0.3491078 0.188807 0.6455073"},{"path":"/articles/advanced.html","id":"sensitivity-analyses-delta-adjustments-and-tipping-point-analyses","dir":"Articles","previous_headings":"","what":"Sensitivity analyses: Delta adjustments and tipping point analyses","title":"rbmi: Advanced Functionality","text":"Delta-adjustments used impute missing data missing random (NMAR) assumption. reflects belief unobserved outcomes systematically “worse” (“better”) “comparable” observed outcomes. extensive discussion delta-adjustment methods, refer Cro et al. (2020). rbmi, marginal delta-adjustment approach implemented. means delta-adjustment applied dataset data imputation MAR reference-based missing data assumptions prior analysis imputed data. Sensitivity analysis using delta-adjustments can therefore performed without re-fit imputation model. rbmi, implemented via delta argument analyse() function.","code":""},{"path":"/articles/advanced.html","id":"simple-delta-adjustments-and-tipping-point-analyses","dir":"Articles","previous_headings":"8 Sensitivity analyses: Delta adjustments and tipping point analyses","what":"Simple delta adjustments and tipping point analyses","title":"rbmi: Advanced Functionality","text":"delta argument analyse() allows users modify outcome variable prior analysis. , user needs provide data.frame contains columns subject visit (identify observation adjusted) plus additional column called delta specifies value added outcomes prior analysis. delta_template() function supports user creating data.frame: creates skeleton data.frame containing one row per subject visit value delta set 0 observations: Note output delta_template() contains additional information can used properly re-set variable delta. example, assume user wants implement delta-adjustment imputed values CIR described section 3. Specifically, assume fixed “worsening adjustment” +5 points applied imputed values regardless treatment group. programmed follows: approach can used implement tipping point analysis. , apply different delta-adjustments imputed data control intervention group, respectively. Assume delta-adjustments less -5 points +15 points considered implausible clinical perspective. Therefore, vary delta-values group -5 +15 points investigate delta combinations lead “tipping” primary analysis result, defined analysis p-value \\(\\geq 0.05\\). According analysis, significant test result primary analysis CIR tipped non-significant result rather extreme delta-adjustments. Please note real analysis recommended use smaller step size grid used .","code":"dat_delta <- delta_template(imputations = impute_obj_CIR) head(dat_delta) #> id visit group is_mar is_missing is_post_ice strategy delta #> 1 id_1 1 Control TRUE TRUE TRUE MAR 0 #> 2 id_1 2 Control TRUE TRUE TRUE MAR 0 #> 3 id_1 3 Control TRUE TRUE TRUE MAR 0 #> 4 id_1 4 Control TRUE TRUE TRUE MAR 0 #> 5 id_1 5 Control TRUE TRUE TRUE MAR 0 #> 6 id_1 6 Control TRUE TRUE TRUE MAR 0 # Set delta-value to 5 for all imputed (previously missing) outcomes and 0 for all other outcomes dat_delta <- delta_template(imputations = impute_obj_CIR) %>% mutate(delta = is_missing * 5) # Repeat the analyses with the delta-adjusted values and pool results ana_delta <- analyse( impute_obj_CIR, delta = dat_delta, vars = vars_an ) pool(ana_delta) #> #> Pool Object #> ----------- #> Number of Results Combined: 20 #> Method: rubin #> Confidence Level: 0.95 #> Alternative: two.sided #> #> Results: #> #> ================================================== #> parameter est se lci uci pval #> -------------------------------------------------- #> trt_1 -0.482 0.524 -1.516 0.552 0.359 #> lsm_ref_1 2.718 0.37 1.987 3.448 <0.001 #> lsm_alt_1 2.235 0.37 1.505 2.966 <0.001 #> trt_2 -0.016 0.56 -1.12 1.089 0.978 #> lsm_ref_2 3.907 0.396 3.125 4.688 <0.001 #> lsm_alt_2 3.891 0.395 3.111 4.671 <0.001 #> trt_3 -1.684 0.641 -2.948 -0.42 0.009 #> lsm_ref_3 6.092 0.452 5.201 6.983 <0.001 #> lsm_alt_3 4.408 0.452 3.515 5.3 <0.001 #> trt_4 -2.359 0.741 -3.821 -0.897 0.002 #> lsm_ref_4 7.951 0.526 6.913 8.99 <0.001 #> lsm_alt_4 5.593 0.522 4.563 6.623 <0.001 #> trt_5 -3.34 0.919 -5.153 -1.526 <0.001 #> lsm_ref_5 9.899 0.643 8.631 11.168 <0.001 #> lsm_alt_5 6.559 0.653 5.271 7.848 <0.001 #> trt_6 -4.21 1.026 -6.236 -2.184 <0.001 #> lsm_ref_6 11.435 0.718 10.019 12.851 <0.001 #> lsm_alt_6 7.225 0.725 5.793 8.656 <0.001 #> -------------------------------------------------- perform_tipp_analysis <- function(delta_control, delta_intervention) { # Derive delta offset based on control and intervention specific deltas delta_df <- delta_df_init %>% mutate( delta_ctl = (group == \"Control\") * is_missing * delta_control, delta_int = (group == \"Intervention\") * is_missing * delta_intervention, delta = delta_ctl + delta_int ) ana_delta <- analyse( impute_obj_CIR, fun = compare_change_lastvisit, vars = vars_an, delta = delta_df, ) pool_delta <- as.data.frame(pool(ana_delta)) list( trt_effect_6 = pool_delta[[\"est\"]], pval_6 = pool_delta[[\"pval\"]] ) } # Get initial delta template delta_df_init <- delta_template(impute_obj_CIR) tipp_frame_grid <- expand.grid( delta_control = seq(-5, 15, by = 2), delta_intervention = seq(-5, 15, by = 2) ) %>% as_tibble() tipp_frame <- tipp_frame_grid %>% mutate( results_list = map2(delta_control, delta_intervention, perform_tipp_analysis), trt_effect_6 = map_dbl(results_list, \"trt_effect_6\"), pval_6 = map_dbl(results_list, \"pval_6\") ) %>% select(-results_list) %>% mutate( pval = cut( pval_6, c(0, 0.001, 0.01, 0.05, 0.2, 1), right = FALSE, labels = c(\"<0.001\", \"0.001 - <0.01\", \"0.01- <0.05\", \"0.05 - <0.20\", \">= 0.20\") ) ) # Show delta values which lead to non-significant analysis results tipp_frame %>% filter(pval_6 >= 0.05) #> # A tibble: 3 × 5 #> delta_control delta_intervention trt_effect_6 pval_6 pval #> #> 1 -5 15 -1.99 0.0935 0.05 - <0.20 #> 2 -3 15 -2.15 0.0704 0.05 - <0.20 #> 3 -1 15 -2.31 0.0527 0.05 - <0.20 ggplot(tipp_frame, aes(delta_control, delta_intervention, fill = pval)) + geom_raster() + scale_fill_manual(values = c(\"darkgreen\", \"lightgreen\", \"lightyellow\", \"orange\", \"red\"))"},{"path":"/articles/advanced.html","id":"more-flexible-delta-adjustments-using-the-dlag-and-delta-arguments-of-delta_template","dir":"Articles","previous_headings":"8 Sensitivity analyses: Delta adjustments and tipping point analyses","what":"More flexible delta-adjustments using the dlag and delta arguments of delta_template()","title":"rbmi: Advanced Functionality","text":"far, discussed simple delta arguments add value imputed values. However, user may want apply flexible delta-adjustments missing values intercurrent event (ICE) vary magnitude delta adjustment depending far away visit question ICE visit. facilitate creation flexible delta-adjustments, delta_template() function two optional additional arguments delta dlag. delta argument specifies default amount delta applied post-ICE visit, whilst dlag specifies scaling coefficient applied based upon visits proximity first visit affected ICE. default, delta added unobserved (.e. imputed) post-ICE outcomes can changed setting optional argument missing_only = FALSE. usage delta dlag arguments best illustrated examples: Assume setting 4 visits user specified delta = c(5,6,7,8) dlag=c(1,2,3,4). subject first visit affected ICE visit 2, values delta dlag imply following delta offset: , subject delta offset 0 applied visit v1, 6 visit v2, 20 visit v3 44 visit v4. Assume instead, subject’s first visit affected ICE visit 3. , values delta dlag imply following delta offset: apply constant delta value +5 visits affected ICE regardless proximity first ICE visit, one set delta = c(5,5,5,5) dlag = c(1,0,0,0). Alternatively, may straightforward setting call delta_template() function without delta dlag arguments overwrite delta column resulting data.frame described previous section (additionally relying is_post_ice variable). Another way using arguments set delta difference time visits dlag amount delta per unit time. example, let’s say visits occur weeks 1, 5, 6 9 want delta 3 applied week ICE. simplicity, assume ICE occurs immediately subject’s last visit affected ICE. achieved setting delta = c(1,4,1,3) (difference weeks visit) dlag = c(3, 3, 3, 3). Assume subject’s first visit affected ICE visit v2, values delta dlag imply following delta offsets: wrap , show action simulated dataset section 2 imputed datasets based CIR assumption section 3. simulation setting specified follow-visits months 2, 4, 6, 8, 10, 12. Assume want apply delta-adjustment 1 every month ICE unobserved post-ICE visits intervention group . (E.g. ICE occurred immediately month 4 visit, total delta applied missing value month 10 visit 6.) program , first use delta dlag arguments delta_template() set corresponding template data.frame: Next, can use additional metadata variables provided delta_template() manually reset delta values control group back 0: Finally, can use delta data.frame apply desired delta offset analysis:","code":"v1 v2 v3 v4 -------------- 5 6 7 8 # delta assigned to each visit 0 1 2 3 # scaling starting from the first visit after the subjects ICE -------------- 0 6 14 24 # delta * scaling -------------- 0 6 20 44 # cumulative sum (i.e. delta) to be applied to each visit v1 v2 v3 v4 -------------- 5 6 7 8 # delta assigned to each visit 0 0 1 2 # scaling starting from the first visit after the subjects ICE -------------- 0 0 7 16 # delta * scaling -------------- 0 0 7 23 # cumulative sum (i.e. delta) to be applied to each visit v1 v2 v3 v4 -------------- 1 4 1 3 # delta assigned to each visit 0 3 3 3 # scaling starting from the first visit after the subjects ICE -------------- 0 12 3 9 # delta * scaling -------------- 0 12 15 24 # cumulative sum (i.e. delta) to be applied to each visit delta_df <- delta_template( impute_obj_CIR, delta = c(2, 2, 2, 2, 2, 2), dlag = c(1, 1, 1, 1, 1, 1) ) head(delta_df) #> id visit group is_mar is_missing is_post_ice strategy delta #> 1 id_1 1 Control TRUE TRUE TRUE MAR 2 #> 2 id_1 2 Control TRUE TRUE TRUE MAR 4 #> 3 id_1 3 Control TRUE TRUE TRUE MAR 6 #> 4 id_1 4 Control TRUE TRUE TRUE MAR 8 #> 5 id_1 5 Control TRUE TRUE TRUE MAR 10 #> 6 id_1 6 Control TRUE TRUE TRUE MAR 12 delta_df2 <- delta_df %>% mutate(delta = if_else(group == \"Control\", 0, delta)) head(delta_df2) #> id visit group is_mar is_missing is_post_ice strategy delta #> 1 id_1 1 Control TRUE TRUE TRUE MAR 0 #> 2 id_1 2 Control TRUE TRUE TRUE MAR 0 #> 3 id_1 3 Control TRUE TRUE TRUE MAR 0 #> 4 id_1 4 Control TRUE TRUE TRUE MAR 0 #> 5 id_1 5 Control TRUE TRUE TRUE MAR 0 #> 6 id_1 6 Control TRUE TRUE TRUE MAR 0 ana_delta <- analyse(impute_obj_CIR, delta = delta_df2, vars = vars_an) pool(ana_delta) #> #> Pool Object #> ----------- #> Number of Results Combined: 20 #> Method: rubin #> Confidence Level: 0.95 #> Alternative: two.sided #> #> Results: #> #> ================================================== #> parameter est se lci uci pval #> -------------------------------------------------- #> trt_1 -0.446 0.514 -1.459 0.567 0.386 #> lsm_ref_1 2.62 0.363 1.904 3.335 <0.001 #> lsm_alt_1 2.173 0.363 1.458 2.889 <0.001 #> trt_2 0.072 0.546 -1.006 1.15 0.895 #> lsm_ref_2 3.708 0.387 2.945 4.471 <0.001 #> lsm_alt_2 3.78 0.386 3.018 4.542 <0.001 #> trt_3 -1.507 0.626 -2.743 -0.272 0.017 #> lsm_ref_3 5.844 0.441 4.973 6.714 <0.001 #> lsm_alt_3 4.336 0.442 3.464 5.209 <0.001 #> trt_4 -2.062 0.731 -3.504 -0.621 0.005 #> lsm_ref_4 7.658 0.519 6.634 8.682 <0.001 #> lsm_alt_4 5.596 0.515 4.58 6.612 <0.001 #> trt_5 -2.938 0.916 -4.746 -1.13 0.002 #> lsm_ref_5 9.558 0.641 8.293 10.823 <0.001 #> lsm_alt_5 6.62 0.651 5.335 7.905 <0.001 #> trt_6 -3.53 1.045 -5.591 -1.469 0.001 #> lsm_ref_6 11.045 0.73 9.604 12.486 <0.001 #> lsm_alt_6 7.515 0.738 6.058 8.971 <0.001 #> --------------------------------------------------"},{"path":[]},{"path":"/articles/quickstart.html","id":"introduction","dir":"Articles","previous_headings":"","what":"Introduction","title":"rbmi: Quickstart","text":"purpose vignette provide 15 minute quickstart guide core functions rbmi package. rbmi package consists 4 core functions (plus several helper functions) typically called sequence: draws() - fits imputation models stores parameters impute() - creates multiple imputed datasets analyse() - analyses multiple imputed datasets pool() - combines analysis results across imputed datasets single statistic","code":""},{"path":"/articles/quickstart.html","id":"the-data","dir":"Articles","previous_headings":"","what":"The Data","title":"rbmi: Quickstart","text":"use publicly available example dataset antidepressant clinical trial active drug versus placebo. relevant endpoint Hamilton 17-item depression rating scale (HAMD17) assessed baseline weeks 1, 2, 4, 6. Study drug discontinuation occurred 24% subjects active drug 26% subjects placebo. data study drug discontinuation missing single additional intermittent missing observation. consider imputation model mean change baseline HAMD17 score outcome (variable CHANGE dataset). following covariates included imputation model: treatment group (THERAPY), (categorical) visit (VISIT), treatment--visit interactions, baseline HAMD17 score (BASVAL), baseline HAMD17 score--visit interactions. common unstructured covariance matrix structure assumed groups. analysis model ANCOVA model treatment group primary factor adjustment baseline HAMD17 score. rbmi expects input dataset complete; , must one row per subject visit. Missing outcome values coded NA, missing covariate values allowed. dataset incomplete, expand_locf() helper function can used add missing rows, using LOCF imputation carry forward observed baseline covariate values visits missing outcomes. Rows corresponding missing outcomes present antidepressant trial dataset. address therefore use expand_locf() function follows:","code":"library(rbmi) library(dplyr) #> #> Attaching package: 'dplyr' #> The following objects are masked from 'package:stats': #> #> filter, lag #> The following objects are masked from 'package:base': #> #> intersect, setdiff, setequal, union data(\"antidepressant_data\") dat <- antidepressant_data # Use expand_locf to add rows corresponding to visits with missing outcomes to the dataset dat <- expand_locf( dat, PATIENT = levels(dat$PATIENT), # expand by PATIENT and VISIT VISIT = levels(dat$VISIT), vars = c(\"BASVAL\", \"THERAPY\"), # fill with LOCF BASVAL and THERAPY group = c(\"PATIENT\"), order = c(\"PATIENT\", \"VISIT\") )"},{"path":"/articles/quickstart.html","id":"draws","dir":"Articles","previous_headings":"","what":"Draws","title":"rbmi: Quickstart","text":"draws() function fits imputation models stores corresponding parameter estimates Bayesian posterior parameter draws. three main inputs draws() function : data - primary longitudinal data.frame containing outcome variable covariates. data_ice - data.frame specifies first visit affected intercurrent event (ICE) imputation strategy handling missing outcome data ICE. one ICE imputed non-MAR strategy allowed per subject. method - statistical method used fit imputation models create imputed datasets. antidepressant trial data, dataset data_ice provided. However, can derived , dataset, subject’s first visit affected ICE “study drug discontinuation” corresponds first terminal missing observation. first derive dateset data_ice create 150 Bayesian posterior draws imputation model parameters. example, assume imputation strategy ICE Jump Reference (JR) subjects 150 multiple imputed datasets using Bayesian posterior draws imputation model created. Note use set_vars() specifies names key variables within dataset imputation model. Additionally, note whilst vars$group vars$visit added terms imputation model default, interaction , thus inclusion group * visit list covariates. Available imputation methods include: Bayesian multiple imputation - method_bayes() Approximate Bayesian multiple imputation - method_approxbayes() Conditional mean imputation (bootstrap) - method_condmean(type = \"bootstrap\") Conditional mean imputation (jackknife) - method_condmean(type = \"jackknife\") Bootstrapped multiple imputation - method = method_bmlmi() comparison methods, refer stat_specs vignette (Section 3.10). “statistical specifications” vignette (Section 3.10): vignette(\"stat_specs\",package=\"rbmi\"). Available imputation strategies include: Missing Random - \"MAR\" Jump Reference - \"JR\" Copy Reference - \"CR\" Copy Increments Reference - \"CIR\" Last Mean Carried Forward - \"LMCF\"","code":"# create data_ice and set the imputation strategy to JR for # each patient with at least one missing observation dat_ice <- dat %>% arrange(PATIENT, VISIT) %>% filter(is.na(CHANGE)) %>% group_by(PATIENT) %>% slice(1) %>% ungroup() %>% select(PATIENT, VISIT) %>% mutate(strategy = \"JR\") # In this dataset, subject 3618 has an intermittent missing values which does not correspond # to a study drug discontinuation. We therefore remove this subject from `dat_ice`. # (In the later imputation step, it will automatically be imputed under the default MAR assumption.) dat_ice <- dat_ice[-which(dat_ice$PATIENT == 3618),] dat_ice #> # A tibble: 43 × 3 #> PATIENT VISIT strategy #> #> 1 1513 5 JR #> 2 1514 5 JR #> 3 1517 5 JR #> 4 1804 7 JR #> 5 2104 7 JR #> 6 2118 5 JR #> 7 2218 6 JR #> 8 2230 6 JR #> 9 2721 5 JR #> 10 2729 5 JR #> # ℹ 33 more rows # Define the names of key variables in our dataset and # the covariates included in the imputation model using `set_vars()` # Note that the covariates argument can also include interaction terms vars <- set_vars( outcome = \"CHANGE\", visit = \"VISIT\", subjid = \"PATIENT\", group = \"THERAPY\", covariates = c(\"BASVAL*VISIT\", \"THERAPY*VISIT\") ) # Define which imputation method to use (here: Bayesian multiple imputation with 150 imputed datsets) method <- method_bayes( burn_in = 200, burn_between = 5, n_samples = 150, seed = 675442751 ) # Create samples for the imputation parameters by running the draws() function set.seed(987) drawObj <- draws( data = dat, data_ice = dat_ice, vars = vars, method = method, quiet = TRUE ) drawObj #> #> Draws Object #> ------------ #> Number of Samples: 150 #> Number of Failed Samples: 0 #> Model Formula: CHANGE ~ 1 + THERAPY + VISIT + BASVAL * VISIT + THERAPY * VISIT #> Imputation Type: random #> Method: #> name: Bayes #> burn_in: 200 #> burn_between: 5 #> same_cov: TRUE #> n_samples: 150 #> seed: 675442751"},{"path":"/articles/quickstart.html","id":"impute","dir":"Articles","previous_headings":"","what":"Impute","title":"rbmi: Quickstart","text":"next step use parameters imputation model generate imputed datasets. done via impute() function. function two key inputs: imputation model output draws() reference groups relevant reference-based imputation methods. ’s usage thus: instance, specifying PLACEBO group reference group well DRUG group (standard imputation using reference-based methods). Generally speaking, need see directly interact imputed datasets. However, wish inspect , can extracted imputation object using extract_imputed_dfs() helper function, .e.: Note case method_bayes() method_approxbayes(), imputed datasets correspond random imputations original dataset. method_condmean(), first imputed dataset always correspond completed original dataset containing subjects. method_condmean(type=\"jackknife\"), remaining datasets correspond conditional mean imputations leave-one-subject-datasets, whereas method_condmean(type=\"bootstrap\"), subsequent dataset corresponds conditional mean imputation bootstrapped datasets. method_bmlmi(), imputed datasets correspond sets random imputations bootstrapped datasets.","code":"imputeObj <- impute( drawObj, references = c(\"DRUG\" = \"PLACEBO\", \"PLACEBO\" = \"PLACEBO\") ) imputeObj #> #> Imputation Object #> ----------------- #> Number of Imputed Datasets: 150 #> Fraction of Missing Data (Original Dataset): #> 4: 0% #> 5: 8% #> 6: 13% #> 7: 25% #> References: #> DRUG -> PLACEBO #> PLACEBO -> PLACEBO imputed_dfs <- extract_imputed_dfs(imputeObj) head(imputed_dfs[[10]], 12) # first 12 rows of 10th imputed dataset #> PATIENT HAMATOTL PGIIMP RELDAYS VISIT THERAPY GENDER POOLINV BASVAL #> 1 new_pt_1 21 2 7 4 DRUG F 006 32 #> 2 new_pt_1 19 2 14 5 DRUG F 006 32 #> 3 new_pt_1 21 3 28 6 DRUG F 006 32 #> 4 new_pt_1 17 4 42 7 DRUG F 006 32 #> 5 new_pt_2 18 3 7 4 PLACEBO F 006 14 #> 6 new_pt_2 18 2 15 5 PLACEBO F 006 14 #> 7 new_pt_2 14 3 29 6 PLACEBO F 006 14 #> 8 new_pt_2 8 2 42 7 PLACEBO F 006 14 #> 9 new_pt_3 18 3 7 4 DRUG F 006 21 #> 10 new_pt_3 17 3 14 5 DRUG F 006 21 #> 11 new_pt_3 12 3 28 6 DRUG F 006 21 #> 12 new_pt_3 9 3 44 7 DRUG F 006 21 #> HAMDTL17 CHANGE #> 1 21 -11 #> 2 20 -12 #> 3 19 -13 #> 4 17 -15 #> 5 11 -3 #> 6 14 0 #> 7 9 -5 #> 8 5 -9 #> 9 20 -1 #> 10 18 -3 #> 11 16 -5 #> 12 13 -8"},{"path":"/articles/quickstart.html","id":"analyse","dir":"Articles","previous_headings":"","what":"Analyse","title":"rbmi: Quickstart","text":"next step run analysis model imputed dataset. done defining analysis function calling analyse() apply function imputed dataset. vignette use ancova() function provided rbmi package fits separate ANCOVA model outcomes visit returns treatment effect estimate corresponding least square means group per visit. Note , similar draws(), ancova() function uses set_vars() function determines names key variables within data covariates (addition treatment group) analysis model adjusted. Please also note names analysis estimates contain “ref” “alt” refer two treatment arms. particular “ref” refers first factor level vars$group necessarily coincide control arm. example, since levels(dat[[vars$group]]) = c(\"DRUG\", PLACEBO), results associated “ref” correspond intervention arm, associated “alt” correspond control arm. Additionally, can use delta argument analyse() perform delta adjustments imputed datasets prior analysis. brief, implemented specifying data.frame contains amount adjustment added longitudinal outcome subject visit, .e.  data.frame must contain columns subjid, visit, delta. appreciated carrying procedure potentially tedious, therefore delta_template() helper function provided simplify . particular, delta_template() returns shell data.frame delta-adjustment set 0 patients. Additionally delta_template() adds several meta-variables onto shell data.frame can used manual derivation manipulation delta-adjustment. example lets say want add delta-value 5 imputed values (.e. values missing original dataset) drug arm. implemented follows:","code":"anaObj <- analyse( imputeObj, ancova, vars = set_vars( subjid = \"PATIENT\", outcome = \"CHANGE\", visit = \"VISIT\", group = \"THERAPY\", covariates = c(\"BASVAL\") ) ) anaObj #> #> Analysis Object #> --------------- #> Number of Results: 150 #> Analysis Function: ancova #> Delta Applied: FALSE #> Analysis Estimates: #> trt_4 #> lsm_ref_4 #> lsm_alt_4 #> trt_5 #> lsm_ref_5 #> lsm_alt_5 #> trt_6 #> lsm_ref_6 #> lsm_alt_6 #> trt_7 #> lsm_ref_7 #> lsm_alt_7 # For reference show the additional meta variables provided delta_template(imputeObj) %>% as_tibble() #> # A tibble: 688 × 8 #> PATIENT VISIT THERAPY is_mar is_missing is_post_ice strategy delta #> #> 1 1503 4 DRUG TRUE FALSE FALSE NA 0 #> 2 1503 5 DRUG TRUE FALSE FALSE NA 0 #> 3 1503 6 DRUG TRUE FALSE FALSE NA 0 #> 4 1503 7 DRUG TRUE FALSE FALSE NA 0 #> 5 1507 4 PLACEBO TRUE FALSE FALSE NA 0 #> 6 1507 5 PLACEBO TRUE FALSE FALSE NA 0 #> 7 1507 6 PLACEBO TRUE FALSE FALSE NA 0 #> 8 1507 7 PLACEBO TRUE FALSE FALSE NA 0 #> 9 1509 4 DRUG TRUE FALSE FALSE NA 0 #> 10 1509 5 DRUG TRUE FALSE FALSE NA 0 #> # ℹ 678 more rows delta_df <- delta_template(imputeObj) %>% as_tibble() %>% mutate(delta = if_else(THERAPY == \"DRUG\" & is_missing , 5, 0)) %>% select(PATIENT, VISIT, delta) delta_df #> # A tibble: 688 × 3 #> PATIENT VISIT delta #> #> 1 1503 4 0 #> 2 1503 5 0 #> 3 1503 6 0 #> 4 1503 7 0 #> 5 1507 4 0 #> 6 1507 5 0 #> 7 1507 6 0 #> 8 1507 7 0 #> 9 1509 4 0 #> 10 1509 5 0 #> # ℹ 678 more rows anaObj_delta <- analyse( imputeObj, ancova, delta = delta_df, vars = set_vars( subjid = \"PATIENT\", outcome = \"CHANGE\", visit = \"VISIT\", group = \"THERAPY\", covariates = c(\"BASVAL\") ) )"},{"path":"/articles/quickstart.html","id":"pool","dir":"Articles","previous_headings":"","what":"Pool","title":"rbmi: Quickstart","text":"Finally, pool() function can used summarise analysis results across multiple imputed datasets provide overall statistic standard error, confidence intervals p-value hypothesis test null hypothesis effect equal 0. Note pooling method automatically derived based method specified original call draws(): method_bayes() method_approxbayes() pooling inference based Rubin’s rules. method_condmean(type = \"bootstrap\") inference either based normal approximation using bootstrap standard error (pool(..., type = \"normal\")) bootstrap percentiles (pool(..., type = \"percentile\")). method_condmean(type = \"jackknife\") inference based normal approximation using jackknife estimate standard error. method = method_bmlmi() inference according methods described von Hippel Bartlett (see stat_specs vignette details) Since used Bayesian multiple imputation vignette, pool() function automatically use Rubin’s rules. table values shown print message poolObj can also extracted using .data.frame() function: outputs gives estimated difference 2.077 (95% CI -0.133 4.287) two groups last visit associated p-value 0.065.","code":"poolObj <- pool( anaObj, conf.level = 0.95, alternative = \"two.sided\" ) poolObj #> #> Pool Object #> ----------- #> Number of Results Combined: 150 #> Method: rubin #> Confidence Level: 0.95 #> Alternative: two.sided #> #> Results: #> #> ================================================== #> parameter est se lci uci pval #> -------------------------------------------------- #> trt_4 -0.092 0.683 -1.439 1.256 0.893 #> lsm_ref_4 -1.616 0.486 -2.576 -0.656 0.001 #> lsm_alt_4 -1.708 0.475 -2.645 -0.77 <0.001 #> trt_5 1.278 0.927 -0.553 3.108 0.17 #> lsm_ref_5 -4.109 0.661 -5.415 -2.804 <0.001 #> lsm_alt_5 -2.831 0.647 -4.109 -1.553 <0.001 #> trt_6 1.907 0.997 -0.062 3.876 0.058 #> lsm_ref_6 -6.084 0.711 -7.49 -4.679 <0.001 #> lsm_alt_6 -4.177 0.698 -5.556 -2.799 <0.001 #> trt_7 2.077 1.118 -0.133 4.287 0.065 #> lsm_ref_7 -6.951 0.81 -8.552 -5.351 <0.001 #> lsm_alt_7 -4.875 0.796 -6.449 -3.3 <0.001 #> -------------------------------------------------- as.data.frame(poolObj) #> parameter est se lci uci pval #> 1 trt_4 -0.09180645 0.6826279 -1.43949684 1.2558839 8.931772e-01 #> 2 lsm_ref_4 -1.61581996 0.4862316 -2.57577141 -0.6558685 1.093708e-03 #> 3 lsm_alt_4 -1.70762640 0.4749573 -2.64531931 -0.7699335 4.262148e-04 #> 4 trt_5 1.27784334 0.9267958 -0.55262591 3.1083126 1.699077e-01 #> 5 lsm_ref_5 -4.10912984 0.6609343 -5.41453523 -2.8037244 4.327320e-09 #> 6 lsm_alt_5 -2.83128650 0.6469374 -4.10909149 -1.5534815 2.189243e-05 #> 7 trt_6 1.90697590 0.9965203 -0.06173754 3.8756893 5.753387e-02 #> 8 lsm_ref_6 -6.08439116 0.7114036 -7.48988902 -4.6788933 1.195687e-14 #> 9 lsm_alt_6 -4.17741526 0.6975145 -5.55556852 -2.7992620 1.482787e-08 #> 10 trt_7 2.07666647 1.1181891 -0.13345442 4.2867874 6.532284e-02 #> 11 lsm_ref_7 -6.95143063 0.8095397 -8.55204130 -5.3508200 1.622026e-14 #> 12 lsm_alt_7 -4.87476416 0.7963105 -6.44945627 -3.3000720 9.230722e-09"},{"path":"/articles/quickstart.html","id":"code","dir":"Articles","previous_headings":"","what":"Code","title":"rbmi: Quickstart","text":"report code presented vignette.","code":"library(rbmi) library(dplyr) data(\"antidepressant_data\") dat <- antidepressant_data # Use expand_locf to add rows corresponding to visits with missing outcomes to the dataset dat <- expand_locf( dat, PATIENT = levels(dat$PATIENT), # expand by PATIENT and VISIT VISIT = levels(dat$VISIT), vars = c(\"BASVAL\", \"THERAPY\"), # fill with LOCF BASVAL and THERAPY group = c(\"PATIENT\"), order = c(\"PATIENT\", \"VISIT\") ) # Create data_ice and set the imputation strategy to JR for # each patient with at least one missing observation dat_ice <- dat %>% arrange(PATIENT, VISIT) %>% filter(is.na(CHANGE)) %>% group_by(PATIENT) %>% slice(1) %>% ungroup() %>% select(PATIENT, VISIT) %>% mutate(strategy = \"JR\") # In this dataset, subject 3618 has an intermittent missing values which does not correspond # to a study drug discontinuation. We therefore remove this subject from `dat_ice`. # (In the later imputation step, it will automatically be imputed under the default MAR assumption.) dat_ice <- dat_ice[-which(dat_ice$PATIENT == 3618),] # Define the names of key variables in our dataset using `set_vars()` # and the covariates included in the imputation model # Note that the covariates argument can also include interaction terms vars <- set_vars( outcome = \"CHANGE\", visit = \"VISIT\", subjid = \"PATIENT\", group = \"THERAPY\", covariates = c(\"BASVAL*VISIT\", \"THERAPY*VISIT\") ) # Define which imputation method to use (here: Bayesian multiple imputation with 150 imputed datsets) method <- method_bayes( burn_in = 200, burn_between = 5, n_samples = 150, seed = 675442751 ) # Create samples for the imputation parameters by running the draws() function set.seed(987) drawObj <- draws( data = dat, data_ice = dat_ice, vars = vars, method = method, quiet = TRUE ) # Impute the data imputeObj <- impute( drawObj, references = c(\"DRUG\" = \"PLACEBO\", \"PLACEBO\" = \"PLACEBO\") ) # Fit the analysis model on each imputed dataset anaObj <- analyse( imputeObj, ancova, vars = set_vars( subjid = \"PATIENT\", outcome = \"CHANGE\", visit = \"VISIT\", group = \"THERAPY\", covariates = c(\"BASVAL\") ) ) # Apply a delta adjustment # Add a delta-value of 5 to all imputed values (i.e. those values # which were missing in the original dataset) in the drug arm. delta_df <- delta_template(imputeObj) %>% as_tibble() %>% mutate(delta = if_else(THERAPY == \"DRUG\" & is_missing , 5, 0)) %>% select(PATIENT, VISIT, delta) # Repeat the analyses with the adjusted values anaObj_delta <- analyse( imputeObj, ancova, delta = delta_df, vars = set_vars( subjid = \"PATIENT\", outcome = \"CHANGE\", visit = \"VISIT\", group = \"THERAPY\", covariates = c(\"BASVAL\") ) ) # Pool the results poolObj <- pool( anaObj, conf.level = 0.95, alternative = \"two.sided\" )"},{"path":"/articles/stat_specs.html","id":"scope-of-this-document","dir":"Articles","previous_headings":"","what":"Scope of this document","title":"rbmi: Statistical Specifications","text":"document describes statistical methods implemented rbmi R package standard reference-based multiple imputation continuous longitudinal outcomes. package implements three classes multiple imputation (MI) approaches: Conventional MI methods based Bayesian (approximate Bayesian) posterior draws model parameters combined Rubin’s rules make inferences described Carpenter, Roger, Kenward (2013) Cro et al. (2020). Conditional mean imputation methods combined re-sampling techniques described Wolbers et al. (2022). Bootstrapped MI methods described von Hippel Bartlett (2021). document structured follows: first provide informal introduction estimands corresponding treatment effect estimation based MI (section 2). core document consists section 3 describes statistical methodology detail also contains comparison implemented approaches (section 3.10). link theory functions included package rbmi described section 4. conclude comparison package alternative software implementations reference-based imputation methods (section 5).","code":""},{"path":[]},{"path":"/articles/stat_specs.html","id":"estimands","dir":"Articles","previous_headings":"2 Introduction to estimands and estimation methods","what":"Estimands","title":"rbmi: Statistical Specifications","text":"ICH E9(R1) addendum estimands sensitivity analyses describes systematic approach ensure alignment among clinical trial objectives, trial execution/conduct, statistical analyses, interpretation results (ICH E9 working group (2019)). per addendum, estimand precise description treatment effect reflecting clinical question posed trial objective summarizes population-level outcomes patients different treatment conditions compared. One important attribute estimand list possible intercurrent events (ICEs), .e. events occurring treatment initiation affect either interpretation existence measurements associated clinical question interest, definition appropriate strategies deal ICEs. three relevant strategies purpose document hypothetical strategy, treatment policy strategy, composite strategy. hypothetical strategy, scenario envisaged ICE occur. scenario, endpoint values ICE directly observable treated using models missing data. treatment policy strategy, treatment effect presence ICEs targeted analyses based observed outcomes regardless whether subject ICE . composite strategy, ICE included component endpoint.","code":""},{"path":"/articles/stat_specs.html","id":"alignment-between-the-estimand-and-the-estimation-method","dir":"Articles","previous_headings":"2 Introduction to estimands and estimation methods","what":"Alignment between the estimand and the estimation method","title":"rbmi: Statistical Specifications","text":"ICH E9(R1) addendum distinguishes ICEs missing data (ICH E9 working group (2019)). Whereas ICEs treatment discontinuations reflect clinical practice, amount missing data can minimized conduct clinical trial. However, many connections missing data ICEs. example, often difficult retain subjects clinical trial treatment discontinuation subject’s dropout trial leads missing data. another example, outcome values ICEs addressed using hypothetical strateg directly observable hypothetical scenario. Consequently, observed outcome values ICEs typically discarded treated missing data. addendum proposes estimation methods address problem presented missing data selected align estimand. recent overview methods align estimator estimand Mallinckrodt et al. (2020). short introduction estimation methods studies longitudinal endpoints can also found Wolbers et al. (2022). One prominent statistical method purpose multiple imputation (MI), target rbmi package.","code":""},{"path":"/articles/stat_specs.html","id":"missing-data-prior-to-ices","dir":"Articles","previous_headings":"2 Introduction to estimands and estimation methods > 2.2 Alignment between the estimand and the estimation method","what":"Missing data prior to ICEs","title":"rbmi: Statistical Specifications","text":"Missing data may occur subjects without ICE prior occurrence ICE. missing outcomes associated ICE, often plausible impute missing--random (MAR) assumption using standard MMRM imputation model longitudinal outcomes. Informally, MAR occurs missing data can fully accounted baseline variables included model observed longitudinal outcomes, model correctly specified.","code":""},{"path":"/articles/stat_specs.html","id":"implementation-of-the-hypothetical-strategy","dir":"Articles","previous_headings":"2 Introduction to estimands and estimation methods > 2.2 Alignment between the estimand and the estimation method","what":"Implementation of the hypothetical strategy","title":"rbmi: Statistical Specifications","text":"MAR imputation model described often also good starting point imputing data ICE handled using hypothetical strategy (Mallinckrodt et al. (2020)). Informally, assumes unobserved values ICE similar observed data subjects ICE remained follow-. However, situations, may reasonable assume missingness “informative” indicates systematically better worse outcome observed subjects. situations, MNAR imputation \\(\\delta\\)-adjustment explored sensitivity analysis. \\(\\delta\\)-adjustments add fixed random quantity imputations order make imputed outcomes systematically worse better observed described Cro et al. (2020). rbmi fixed \\(\\delta\\)-adjustments implemented.","code":""},{"path":"/articles/stat_specs.html","id":"implementation-of-the-treatment-policy-strategy","dir":"Articles","previous_headings":"2 Introduction to estimands and estimation methods > 2.2 Alignment between the estimand and the estimation method","what":"Implementation of the treatment policy strategy","title":"rbmi: Statistical Specifications","text":"Ideally, data collection continues ICE handled treatment policy strategy missing data arises. Indeed, post-ICE data increasingly systematically collected RCTs. However, despite best efforts, missing data ICE study treatment discontinuation may still occur subject drops study discontinuation. difficult give definite recommendations regarding implementation treatment policy strategy presence missing data stage optimal method highly context dependent topic ongoing statistical research. ICEs thought negligible effect efficacy outcomes, standard MAR-based imputation may appropriate. contrast, ICE treatment discontinuation may expected substantial impact efficacy outcomes. settings, MAR assumption may still plausible conditioning subject’s time-varying treatment status (Guizzaro et al. (2021)). case, one option impute missing post-discontinuation data based subjects also discontinued treatment continued followed (Polverejan Dragalin (2020)). Another option may require somewhat less post-discontinuation data include subjects imputation procedure model post-discontinuation data using time-varying treatment status indicators (e.g. time-varying indicators treatment compliance, discontinuation, initiation rescue treatment) (Guizzaro et al. (2021)). approach, post-ICE outcomes included every step analysis, including fitting imputation model. assumes ICEs may impact post-ICE outcomes otherwise missingness non-informative. approach also assumes time-varying covariates contain missing values, deviations outcomes ICE correctly modeled time-varying covariates, sufficient post-ICE data available inform regression coefficients time-varying covariates. proposals relatively recent remain open questions regarding appropriate trade-model complexity (e.g. model account potentially differential effect post-ICE outcomes depending timing ICE?) variance resulting treatment effect estimate. generally, yet established much post-discontinuation data required implement methods robustly without risk substantial inflation variance. trial settings, subjects discontinue randomized treatment. settings, treatment discontinuation rates higher difficult retain subjects trial treatment discontinuation leading sparse data collection treatment discontinuation. settings, amount available data treatment discontinuation may insufficient inform imputation model explicitly models post-discontinuation data. Depending disease area anticipated mechanism action intervention, may plausible assume subjects intervention group behave similarly subjects control group ICE treatment discontinuation. case, reference-based imputation methods option (Mallinckrodt et al. (2020)). Reference-based imputation methods formalize idea impute missing data intervention group based data control reference group. general description review reference-based imputation methods, refer Carpenter, Roger, Kenward (2013), Cro et al. (2020), . White, Royes, Best (2020) Wolbers et al. (2022). technical description implemented statistical methodology reference-based imputation, refer section 3 (particular section 3.4).","code":""},{"path":"/articles/stat_specs.html","id":"implementation-of-the-composite-strategy","dir":"Articles","previous_headings":"2 Introduction to estimands and estimation methods > 2.2 Alignment between the estimand and the estimation method","what":"Implementation of the composite strategy","title":"rbmi: Statistical Specifications","text":"composite strategy typically applied binary time--event outcomes can also used continuous outcomes ascribing suitably unfavorable value patients experience ICEs composite strategy defined. One possibility implement use MI \\(\\delta\\)-adjustment post-ICE data described Darken et al. (2020).","code":""},{"path":[]},{"path":"/articles/stat_specs.html","id":"sec:methodsOverview","dir":"Articles","previous_headings":"3 Statistical methodology","what":"Overview of the imputation procedure","title":"rbmi: Statistical Specifications","text":"Analyses datasets missing data always rely missing data assumptions. methods described can used produce valid imputations MAR assumption reference-based imputation assumptions. MNAR imputation based fixed \\(\\delta\\)-adjustments typically used sensitivity analyses tipping-point analyses also supported. Three general imputation approaches implemented rbmi: Conventional MI based Bayesian (approximate Bayesian) posterior draws imputation model combined Rubin’s rules inference described Carpenter, Roger, Kenward (2013) Cro et al. (2020). Conditional mean imputation based REML estimate imputation model combined resampling techniques (jackknife bootstrap) inference described Wolbers et al. (2022). Bootstrapped MI methods based REML estimates imputation model described von Hippel Bartlett (2021).","code":""},{"path":"/articles/stat_specs.html","id":"conventional-mi","dir":"Articles","previous_headings":"3 Statistical methodology > 3.1 Overview of the imputation procedure","what":"Conventional MI","title":"rbmi: Statistical Specifications","text":"Conventional MI approaches include following steps: Base imputation model fitting step (Section 3.3) Fit Bayesian multivariate normal mixed model repeated measures (MMRM) observed longitudinal outcomes exclusion data ICEs reference-based missing data imputation desired (Section 3.3.3). Draw \\(M\\) posterior samples estimated parameters (regression coefficients covariance matrices) model. Alternatively, \\(M\\) approximate posterior draws posterior distribution can sampled repeatedly applying conventional restricted maximum-likelihood (REML) parameter estimation MMRM model nonparametric bootstrap samples original dataset (Section 3.3.4). Imputation step (Section 3.4) Take single sample \\(m\\) (\\(m\\1,\\ldots, M)\\) posterior distribution imputation model parameters. subject, use sampled parameters defined imputation strategy determine mean covariance matrix describing subject’s marginal outcome distribution longitudinal outcome assessments (.e. observed missing outcomes). subjects, construct conditional multivariate normal distribution missing outcomes given observed outcomes (including observed outcomes ICEs reference-based assumption desired). subject, draw single sample conditional distribution impute missing outcomes leading complete imputed dataset. sensitivity analyses, pre-defined \\(\\delta\\)-adjustment may applied imputed data prior analysis step. (Section 3.5). Analysis step (Section 3.6) Analyze imputed dataset using analysis model (e.g. ANCOVA) resulting point estimate standard error (corresponding degrees freedom) treatment effect. Pooling step inference (Section 3.7) Repeat steps 2. 3. posterior sample \\(m\\), resulting \\(M\\) complete datasets, \\(M\\) point estimates treatment effect, \\(M\\) standard errors (corresponding degrees freedom). Pool \\(M\\) treatment effect estimates, standard errors, degrees freedom using rules Barnard Rubin obtain final pooled treatment effect estimator, standard error, degrees freedom.","code":""},{"path":"/articles/stat_specs.html","id":"conditional-mean-imputation","dir":"Articles","previous_headings":"3 Statistical methodology > 3.1 Overview of the imputation procedure","what":"Conditional mean imputation","title":"rbmi: Statistical Specifications","text":"conditional mean imputation approach includes following steps: Base imputation model fitting step (Section 3.3) Fit conventional multivariate normal/MMRM model using restricted maximum likelihood (REML) observed longitudinal outcomes exclusion data ICEs reference-based missing data imputation desired (Section 3.3.2). Imputation step (Section 3.4) subject, use fitted parameters step 1. construct conditional distribution missing outcomes given observed outcomes (including observed outcomes ICEs reference-based missing data imputation desired) described . subject, impute missing data deterministically mean conditional distribution leading complete imputed dataset. sensitivity analyses, pre-defined \\(\\delta\\)-adjustment may applied imputed data prior analysis step. (Section 3.5). Analysis step (Section 3.6) Apply analysis model (e.g. ANCOVA) completed dataset resulting point estimate treatment effect. Jackknife bootstrap inference step (Section 3.8) Inference treatment effect estimate 3. based re-sampling techniques. jackknife bootstrap supported. Importantly, methods require repeating steps imputation procedure (.e. imputation, conditional mean imputation, analysis steps) resampled datasets.","code":""},{"path":"/articles/stat_specs.html","id":"bootstrapped-mi","dir":"Articles","previous_headings":"3 Statistical methodology > 3.1 Overview of the imputation procedure","what":"Bootstrapped MI","title":"rbmi: Statistical Specifications","text":"bootstrapped MI approach includes following steps: Base imputation model fitting step (Section 3.3) Apply conventional restricted maximum-likelihood (REML) parameter estimation MMRM model \\(B\\) nonparametric bootstrap samples original dataset using observed longitudinal outcomes exclusion data ICEs reference-based missing data imputation desired. Imputation step (Section 3.4) Take bootstrapped dataset \\(b\\) (\\(b\\1,\\ldots, B)\\) corresponding imputation model parameter estimates. subject (bootstrapped dataset), use parameter estimates defined strategy dealing ICEs determine mean covariance matrix describing subject’s marginal outcome distribution longitudinal outcome assessments (.e. observed missing outcomes). subjects (bootstrapped dataset), construct conditional multivariate normal distribution missing outcomes given observed outcomes (including observed outcomes ICEs reference-based missing data imputation desired). subject (bootstrapped dataset), draw \\(D\\) samples conditional distributions impute missing outcomes leading \\(D\\) complete imputed dataset bootstrap sample \\(b\\). sensitivity analyses, pre-defined \\(\\delta\\)-adjustment may applied imputed data prior analysis step. (Section 3.5). Analysis step (Section 3.6) Analyze \\(B\\times D\\) imputed datasets using analysis model (e.g. ANCOVA) resulting \\(B\\times D\\) point estimates treatment effect. Pooling step inference (Section 3.9) Pool \\(B\\times D\\) treatment effect estimates described von Hippel Bartlett (2021) obtain final pooled treatment effect estimate, standard error, degrees freedom.","code":""},{"path":"/articles/stat_specs.html","id":"setting-notation-and-missing-data-assumptions","dir":"Articles","previous_headings":"3 Statistical methodology","what":"Setting, notation, and missing data assumptions","title":"rbmi: Statistical Specifications","text":"Assume data study \\(n\\) subjects total subject \\(\\) (\\(=1,\\ldots,n\\)) \\(J\\) scheduled follow-visits outcome interest assessed. applications, data randomized trial intervention vs control group treatment effect interest comparison outcomes specific visit randomized groups. However, single-arm trials multi-arm trials principle also supported rbmi implementation. Denote observed outcome vector length \\(J\\) subject \\(\\) \\(Y_i\\) (missing assessments coded NA (available)) non-missing missing components \\(Y_{!}\\) \\(Y_{?}\\), respectively. default, imputation missing outcomes \\(Y_{}\\) performed MAR assumption rbmi. Therefore, missing data following ICE handled using MAR imputation, compatible default assumption. discussed Section 2, MAR assumption often good starting point implementing hypothetical strategy. also note observed outcome data ICE handled using hypothetical strategy compatible strategy. Therefore, assume post-ICE data ICEs handled using hypothetical strategy already set NA \\(Y_i\\) prior calling rbmi functions. However, observed outcomes ICEs handled using treatment policy strategy included \\(Y_i\\) compatible strategy. Subjects may also experience one ICE missing data imputation according reference-based imputation method foreseen. subject \\(\\) ICE, denote first visit affected ICE \\(\\tilde{t}_i \\\\{1,\\ldots,J\\}\\). subjects, set \\(\\tilde{t}_i=\\infty\\). subject’s outcome vector setting observed outcomes visit \\(\\tilde{t}_i\\) onwards missing (.e. NA) denoted \\(Y'_i\\) corresponding data vector removal NA elements \\(Y'_{!}\\). MNAR \\(\\delta\\)-adjustments added imputed datasets formal imputation steps. covered separate section (Section 3.5).","code":""},{"path":[]},{"path":"/articles/stat_specs.html","id":"sec:imputationModelSpecs","dir":"Articles","previous_headings":"3 Statistical methodology > 3.3 The base imputation model","what":"Included data and model specification","title":"rbmi: Statistical Specifications","text":"purpose imputation model estimate (covariate-dependent) mean trajectories covariance matrices group absence ICEs handled using reference-based imputation methods. Conventionally, publications reference-based imputation methods implicitly assumed corresponding post-ICE data missing subjects (Carpenter, Roger, Kenward (2013)). also allow situation post-ICE data available subjects needs imputed using reference-based methods others. However, observed data ICEs reference-based imputation methods specified compatible imputation model described therefore removed considered missing purpose estimating imputation model, purpose . example, patient ICE addressed reference-based method outcomes ICE collected, post-ICE outcomes excluded fitting base imputation model (included following steps). , base imputation model fitted \\(Y'_{!}\\) \\(Y_{!}\\). exclude data, imputation model mistakenly estimate mean trajectories based mixture observed pre- post-ICE data relevant reference-based imputations. Observed post-ICE outcomes control reference group also excluded base imputation model user specifies reference-based imputation strategy ICEs. ensures ICE impact data included imputation model regardless whether ICE occurred control intervention group. hand, imputation reference group based MAR assumption even reference-based imputation methods may preferable settings include post-ICE data control group base imputation model. can implemented specifying MAR strategy ICE control group reference-based strategy ICE intervention group. base imputation model longitudinal outcomes \\(Y'_i\\) assumes mean structure linear function covariates. Full flexibility specification linear predictor model supported. minimum covariates include treatment group, (categorical) visit, treatment--visit interactions. Typically, covariates including baseline outcome also included. External time-varying covariates (e.g. calendar time visit) well internal time-varying (e.g. time-varying indicators treatment discontinuation initiation rescue treatment) may principle also included indicated (Guizzaro et al. (2021)). Missing covariate values allowed. means values time-varying covariates must non-missing every visit regardless whether outcome measured missing. Denote \\(J\\times p\\) design matrix subject \\(\\) corresponding mean structure model \\(X_i\\) matrix removal rows corresponding missing outcomes \\(Y'_{!}\\) \\(X'_{!}\\). \\(p\\) number parameters mean structure model elements \\(Y'_{!}\\). base imputation model observed outcomes defined : \\[ Y'_{!} = X'_{!}\\beta + \\epsilon_{!} \\mbox{ } \\epsilon_{!}\\sim N(0,\\Sigma_{!!})\\] \\(\\beta\\) vector regression coefficients \\(\\Sigma_{!!}\\) covariance matrix obtained complete-data \\(J\\times J\\)-covariance matrix \\(\\Sigma\\) omitting rows columns corresponding missing outcome assessments subject \\(\\). Typically, common unstructured covariance matrix subjects assumed \\(\\Sigma\\) separate covariate matrices per treatment group also supported. Indeed, implementation also supports specification separate covariate matrices according arbitrarily defined categorical variable groups subjects disjoint subset. example, useful different covariance matrices suspected different subject strata. Finally, imputation methods described rely Bayesian model fitting MCMC, flexibility choice covariance structure, .e. unstructured (default), heterogeneous Toeplitz, heterogeneous compound symmetry, AR(1) covariance structures supported.","code":""},{"path":"/articles/stat_specs.html","id":"sec:imputationModelREML","dir":"Articles","previous_headings":"3 Statistical methodology > 3.3 The base imputation model","what":"Restricted maximum likelihood estimation (REML)","title":"rbmi: Statistical Specifications","text":"Frequentist parameter estimation base imputation based REML. use REML improved alternative maximum likelihood (ML) covariance parameter estimation originally proposed Patterson Thompson (1971). Since , become default method parameter estimation linear mixed effects models. rbmi allows choose ML REML methods estimate model parameters, REML default option.","code":""},{"path":"/articles/stat_specs.html","id":"sec:imputationModelBayes","dir":"Articles","previous_headings":"3 Statistical methodology > 3.3 The base imputation model","what":"Bayesian model fitting","title":"rbmi: Statistical Specifications","text":"Bayesian imputation model fitted R package rstan (Stan Development Team (2020)). rstan R interface Stan. Stan powerful flexible statistical software developed dedicated team implements Bayesian inference state---art MCMC sampling procedures. multivariate normal model missing data specified section 3.3.1 can considered generalization models described Stan user’s guide (see Stan Development Team (2020, sec. 3.5)). prior distributions SAS implementation “five macros” used (Roger (2021)), .e. improper flat priors regression coefficients weakly informative inverse Wishart prior covariance matrix (matrices). Specifically, let \\(S \\\\mathbb{R}^{J \\times J}\\) symmetric positive definite matrix \\(\\nu \\(J-1, \\infty)\\). symmetric positive definite matrix \\(x \\\\mathbb{R}^{J \\times J}\\) density: \\[ \\text{InvWish}(x \\vert \\nu, S) = \\frac{1}{2^{\\nu J/2}} \\frac{1}{\\Gamma_J(\\frac{\\nu}{2})} \\vert S \\vert^{\\nu/2} \\vert x \\vert ^{-(\\nu + J + 1)/2} \\text{exp}(-\\frac{1}{2} \\text{tr}(Sx^{-1})). \\] \\(\\nu > J+1\\) mean given : \\[ E[x] = \\frac{S}{\\nu - J - 1}. \\] choose \\(S\\) equal estimated covariance matrix frequentist REML fit \\(\\nu = J+2\\) lowest degrees freedom guarantee finite mean. Setting degrees freedom low \\(\\nu\\) ensures prior little impact posterior. Moreover, choice allows interpret parameter \\(S\\) mean prior distribution. “five macros”, MCMC algorithm initialized parameters frequentist REML fit (see section 3.3.2). described , using weakly informative priors parameters. Therefore, Markov chain essentially starting targeted stationary posterior distribution minimal amount burn-chain required.","code":""},{"path":"/articles/stat_specs.html","id":"sec:imputationModelBoot","dir":"Articles","previous_headings":"3 Statistical methodology > 3.3 The base imputation model","what":"Approximate Bayesian posterior draws via the bootstrap","title":"rbmi: Statistical Specifications","text":"Several authors suggested stabler way get Bayesian posterior draws imputation model bootstrap incomplete data calculate REML estimates bootstrap sample (Little Rubin (2002), Efron (1994), Honaker King (2010), von Hippel Bartlett (2021)). method proper REML estimates bootstrap samples asymptotically equivalent sample posterior distribution may provide additional robustness model misspecification (Little Rubin (2002, sec. 10.2.3, part 6), Honaker King (2010)). order retain balance treatment groups stratification factors across bootstrap samples, user able provide stratification variables bootstrap rbmi implementation.","code":""},{"path":[]},{"path":"/articles/stat_specs.html","id":"sec:imputatioMNAR","dir":"Articles","previous_headings":"3 Statistical methodology > 3.4 Imputation step","what":"Marginal imputation distribution for a subject - MAR case","title":"rbmi: Statistical Specifications","text":"subject \\(\\), marginal distribution complete \\(J\\)-dimensional outcome vector assessment visits according imputation model multivariate normal distribution. mean \\(\\tilde{\\mu}_i\\) given predicted mean imputation model conditional subject’s baseline characteristics, group, , optionally, time-varying covariates. covariance matrix \\(\\tilde{\\Sigma}_i\\) given overall estimated covariance matrix , different covariance matrices assumed different groups, covariance matrix corresponding subject \\(\\)’s group.","code":""},{"path":"/articles/stat_specs.html","id":"sec:imputationRefBased","dir":"Articles","previous_headings":"3 Statistical methodology > 3.4 Imputation step","what":"Marginal imputation distribution for a subject - reference-based imputation methods","title":"rbmi: Statistical Specifications","text":"subject \\(\\), calculate mean covariance matrix complete \\(J\\)-dimensional outcome vector assessment visits MAR case denote \\(\\mu_i\\) \\(\\Sigma_i\\). reference-based imputation methods, corresponding reference group also required group. Typically, reference group intervention group control group. reference mean \\(\\mu_{ref,}\\) defined predicted mean imputation model conditional reference group (rather actual group subject \\(\\) belongs ) subject’s baseline characteristics. reference covariance matrix \\(\\Sigma_{ref,}\\) overall estimated covariance matrix , different covariance matrices assumed different groups, estimated covariance matrix corresponding reference group. principle, time-varying covariates also included reference-based imputation methods. However, sensible external time-varying covariates (e.g. calendar time visit) internal time-varying covariates (e.g. treatment discontinuation) latter likely depend actual treatment group typically sensible assume trajectory time-varying covariate reference group. Based means covariance matrices, subject’s marginal imputation distribution reference-based imputation methods calculated detailed Carpenter, Roger, Kenward (2013, sec. 4.3). Denote mean covariance matrix marginal imputation distribution \\(\\tilde{\\mu}_i\\) \\(\\tilde{\\Sigma}_i\\). Recall subject’s first visit affected ICE denoted \\(\\tilde{t}_i \\\\{1,\\ldots,J\\}\\) (visit \\(\\tilde{t}_i-1\\) last visit unaffected ICE). marginal distribution patient \\(\\) built according specific assumption data post ICE follows: Jump reference (JR): patient’s outcome distribution normally distributed following mean: \\[\\tilde{\\mu}_i = (\\mu_i[1], \\dots, \\mu_i[\\tilde{t}_i-1], \\mu_{ref,}[\\tilde{t}_i], \\dots, \\mu_{ref,}[J])^T.\\] covariance matrix constructed follows. First, partition covariance matrices \\(\\Sigma_i\\) \\(\\Sigma_{ref,}\\) blocks according time ICE \\(\\tilde{t}_i\\): \\[ \\Sigma_{} = \\begin{bmatrix} \\Sigma_{, 11} & \\Sigma_{, 12} \\\\ \\Sigma_{, 21} & \\Sigma_{,22} \\\\ \\end{bmatrix} \\] \\[ \\Sigma_{ref,} = \\begin{bmatrix} \\Sigma_{ref, , 11} & \\Sigma_{ref, , 12} \\\\ \\Sigma_{ref, , 21} & \\Sigma_{ref, ,22} \\\\ \\end{bmatrix}. \\] want covariance matrix \\(\\tilde{\\Sigma}_i\\) match \\(\\Sigma_i\\) pre-deviation measurements, \\(\\Sigma_{ref,}\\) conditional components post-deviation given pre-deviation measurements. solution derived Carpenter, Roger, Kenward (2013, sec. 4.3) given : \\[ \\begin{matrix} \\tilde{\\Sigma}_{,11} = \\Sigma_{, 11} \\\\ \\tilde{\\Sigma}_{, 21} = \\Sigma_{ref,, 21} \\Sigma^{-1}_{ref,, 11} \\Sigma_{, 11} \\\\ \\tilde{\\Sigma}_{, 22} = \\Sigma_{ref, , 22} - \\Sigma_{ref,, 21} \\Sigma^{-1}_{ref,, 11} (\\Sigma_{ref,, 11} - \\Sigma_{,11}) \\Sigma^{-1}_{ref,, 11} \\Sigma_{ref,, 12}. \\end{matrix} \\] Copy increments reference (CIR): patient’s outcome distribution normally distributed following mean: \\[ \\begin{split} \\tilde{\\mu}_i =& (\\mu_i[1], \\dots, \\mu_i[\\tilde{t}_i-1], \\mu_i[\\tilde{t}_i-1] + (\\mu_{ref,}[\\tilde{t}_i] - \\mu_{ref,}[\\tilde{t}_i-1]), \\dots,\\\\ & \\mu_i[\\tilde{t}_i-1]+(\\mu_{ref,}[J] - \\mu_{ref,}[\\tilde{t}_i-1]))^T. \\end{split} \\] covariance matrix derived JR method. Copy reference (CR): patient’s outcome distribution normally distributed mean covariance matrix taken reference group: \\[ \\tilde{\\mu}_i = \\mu_{ref,} \\] \\[ \\tilde{\\Sigma}_i = \\Sigma_{ref,}. \\] Last mean carried forward (LMCF): patient’s outcome distribution normally distributed following mean: \\[ \\tilde{\\mu}_i = (\\mu_i[1], \\dots, \\mu_i[\\tilde{t}_i-1], \\mu_i[\\tilde{t}_i-1], \\dots, \\mu_i[\\tilde{t}_i-1])'\\] covariance matrix: \\[ \\tilde{\\Sigma}_i = \\Sigma_i.\\]","code":""},{"path":"/articles/stat_specs.html","id":"sec:imputationRandomConditionalMean","dir":"Articles","previous_headings":"3 Statistical methodology > 3.4 Imputation step","what":"Imputation of missing outcome data","title":"rbmi: Statistical Specifications","text":"joint marginal multivariate normal imputation distribution subject \\(\\)’s observed missing outcome data mean \\(\\tilde{\\mu}_i\\) covariance matrix \\(\\tilde{\\Sigma}_i\\) defined . actual imputation missing outcome data obtained conditioning marginal distribution subject’s observed outcome data. note, approach valid regardless whether subject intermittent terminal missing data. conditional distribution used imputation multivariate normal distribution explicit formulas conditional mean covariance readily available. completeness, report notation terminology setting. marginal distribution outcome patient \\(\\) \\(Y_i \\sim N(\\tilde{\\mu}_i, \\tilde{\\Sigma}_i)\\) outcome \\(Y_i\\) can decomposed observed (\\(Y_{,!}\\)) unobserved (\\(Y_{,?}\\)) components. Analogously mean \\(\\tilde{\\mu}_i\\) can decomposed \\((\\tilde{\\mu}_{,!},\\tilde{\\mu}_{,?})\\) covariance \\(\\tilde{\\Sigma}_i\\) : \\[ \\tilde{\\Sigma}_i = \\begin{bmatrix} \\tilde{\\Sigma}_{, !!} & \\tilde{\\Sigma}_{,!?} \\\\ \\tilde{\\Sigma}_{, ?!} & \\tilde{\\Sigma}_{, ??} \\end{bmatrix}. \\] conditional distribution \\(Y_{,?}\\) conditional \\(Y_{,!}\\) multivariate normal distribution expectation \\[ E(Y_{,?} \\vert Y_{,!})= \\tilde{\\mu}_{,?} + \\tilde{\\Sigma}_{, ?!} \\tilde{\\Sigma}_{,!!}^{-1} (Y_{,!} - \\tilde{\\mu}_{,!}) \\] covariance matrix \\[ Cov(Y_{,?} \\vert Y_{,!}) = \\tilde{\\Sigma}_{,??} - \\tilde{\\Sigma}_{,?!} \\tilde{\\Sigma}_{,!!}^{-1} \\tilde{\\Sigma}_{,!?}. \\] Conventional random imputation consists sampling conditional multivariate normal distribution. Conditional mean imputation imputes missing values deterministic conditional expectation \\(E(Y_{,?} \\vert Y_{,!})\\).","code":""},{"path":"/articles/stat_specs.html","id":"sec:deltaAdjustment","dir":"Articles","previous_headings":"3 Statistical methodology","what":"\\(\\delta\\)-adjustment","title":"rbmi: Statistical Specifications","text":"marginal \\(\\delta\\)-adjustment approach similar “five macros” SAS implemented (Roger (2021)), .e. fixed non-stochastic values added multivariate normal imputation step prior analysis. relevant sensitivity analyses order make imputed data systematically worse better, respectively, observed data. addition, authors suggested \\(\\delta\\)-type adjustments implement composite strategy continuous outcomes (Darken et al. (2020)). implementation provides full flexibility regarding specific implementation \\(\\delta\\)-adjustment, .e. value added may depend randomized treatment group, timing subject’s ICE, factors. suggestions case studies regarding topic, refer Cro et al. (2020).","code":""},{"path":"/articles/stat_specs.html","id":"sec:analysis","dir":"Articles","previous_headings":"3 Statistical methodology","what":"Analysis step","title":"rbmi: Statistical Specifications","text":"data imputation, standard analysis model can applied completed data resulting treatment effect estimate. imputed data longer contains missing values, analysis model often simple. example, can analysis covariance (ANCOVA) model outcome (change outcome baseline) specific visit j dependent variable, randomized treatment group primary covariate , typically, adjustment baseline covariates imputation model.","code":""},{"path":"/articles/stat_specs.html","id":"sec:pooling","dir":"Articles","previous_headings":"3 Statistical methodology","what":"Pooling step for inference of (approximate) Bayesian MI and Rubin’s rules","title":"rbmi: Statistical Specifications","text":"Assume analysis model applied \\(M\\) multiple imputed random datasets resulted \\(m\\) treatment effect estimates \\(\\hat{\\theta}_m\\) (\\(m=1,\\ldots,M\\)) corresponding standard error \\(SE_m\\) (available) degrees freedom \\(\\nu_{com}\\). degrees freedom available analysis model, set \\(\\nu_{com}=\\infty\\) inference based normal distribution. Rubin’s rules used pooling treatment effect estimates corresponding variances estimates analysis steps across \\(M\\) multiple imputed datasets. According Rubin’s rules, final estimate treatment effect calculated sample mean \\(M\\) treatment effect estimates: \\[ \\hat{\\theta} = \\frac{1}{M} \\sum_{m = 1}^M \\hat{\\theta}_m. \\] pooled variance based two components reflect within variance treatment effects across multiple imputed datasets: \\[ V(\\hat{\\theta}) = V_W(\\hat{\\theta}) + (1 + \\frac{1}{M}) V_B(\\hat{\\theta}) \\] \\(V_W(\\hat{\\theta}) = \\frac{1}{M}\\sum_{m = 1}^M SE^2_m\\) within-variance \\(V_B(\\hat{\\theta}) = \\frac{1}{M-1} \\sum_{m = 1}^M (\\hat{\\theta}_m - \\hat{\\theta})^2\\) -variance. Confidence intervals tests null hypothesis \\(H_0: \\theta=\\theta_0\\) based \\(t\\)-statistics \\(T\\): \\[ T= (\\hat{\\theta}-\\theta_0)/\\sqrt{V(\\hat{\\theta})}. \\] null hypothesis, \\(T\\) approximate \\(t\\)-distribution \\(\\nu\\) degrees freedom. \\(\\nu\\) calculated according Barnard Rubin approximation, see Barnard Rubin (1999) (formula 3) Little Rubin (2002) (formula (5.24), page 87): \\[ \\nu = \\frac{\\nu_{old}* \\nu_{obs}}{\\nu_{old} + \\nu_{obs}} \\] \\[ \\nu_{old} = \\frac{M-1}{\\lambda^2} \\quad\\mbox{}\\quad \\nu_{obs} = \\frac{\\nu_{com} + 1}{\\nu_{com} + 3} \\nu_{com} (1 - \\lambda) \\] \\(\\lambda = \\frac{(1 + \\frac{1}{M})V_B(\\hat{\\theta})}{V(\\hat{\\theta})}\\) fraction missing information.","code":""},{"path":[]},{"path":"/articles/stat_specs.html","id":"point-estimate-of-the-treatment-effect","dir":"Articles","previous_headings":"3 Statistical methodology > 3.8 Bootstrap and jackknife inference for conditional mean imputation","what":"Point estimate of the treatment effect","title":"rbmi: Statistical Specifications","text":"point estimator obtained applying analysis model (Section 3.6) single conditional mean imputation missing data (see Section 3.4.3) based REML estimator parameters imputation model (see Section 3.3.2). denote treatment effect estimator \\(\\hat{\\theta}\\). demonstrated Wolbers et al. (2022) (Section 2.4), treatment effect estimator valid analysis model ANCOVA model , generally, treatment effect estimator linear function imputed outcome vector. Indeed, case, estimator identical pooled treatment effect across multiple random REML imputation infinite number imputations corresponds computationally efficient implementation proposal von Hippel Bartlett (2021). expect conditional mean imputation method also applicable analysis models (e.g. general MMRM analysis models) formally justified.","code":""},{"path":"/articles/stat_specs.html","id":"jackknife-standard-errors-confidence-intervals-ci-and-tests-for-the-treatment-effect","dir":"Articles","previous_headings":"3 Statistical methodology > 3.8 Bootstrap and jackknife inference for conditional mean imputation","what":"Jackknife standard errors, confidence intervals (CI) and tests for the treatment effect","title":"rbmi: Statistical Specifications","text":"dataset containing \\(n\\) subjects, jackknife standard error depends treatment effect estimates \\(\\hat{\\theta}_{(-b)}\\) (\\(b=1,\\ldots,n\\)) samples original dataset leave observation subject \\(b\\). described previously, obtain treatment effect estimates leave-one-subject-datasets, steps imputation procedure (.e. imputation, conditional mean imputation, analysis steps) need repeated new dataset. , jackknife standard error defined \\[\\hat{se}_{jack}=[\\frac{(n-1)}{n}\\cdot\\sum_{b=1}^{n} (\\hat{\\theta}_{(-b)}-\\bar{\\theta}_{(.)})^2]^{1/2}\\] \\(\\bar{\\theta}_{(.)}\\) denotes mean jackknife estimates (Efron Tibshirani (1994), chapter 10). corresponding two-sided normal approximation \\(1-\\alpha\\) CI defined \\(\\hat{\\theta}\\pm z^{1-\\alpha/2}\\cdot \\hat{se}_{jack}\\) \\(\\hat{\\theta}\\) treatment effect estimate original dataset. Tests null hypothesis \\(H_0: \\theta=\\theta_0\\) based \\(Z\\)-score \\(Z=(\\hat{\\theta}-\\theta_0)/\\hat{se}_{jack}\\) using standard normal approximation. simulation study reported Wolbers et al. (2022) demonstrated exact protection type error jackknife-based inference relatively low sample size (n = 100 per group) substantial amount missing data (>25% subjects ICE).","code":""},{"path":"/articles/stat_specs.html","id":"bootstrap-standard-errors-confidence-intervals-ci-and-tests-for-the-treatment-effect","dir":"Articles","previous_headings":"3 Statistical methodology > 3.8 Bootstrap and jackknife inference for conditional mean imputation","what":"Bootstrap standard errors, confidence intervals (CI) and tests for the treatment effect","title":"rbmi: Statistical Specifications","text":"alternative jackknife, bootstrap also implemented rbmi (Efron Tibshirani (1994), Davison Hinkley (1997)). Two different bootstrap methods implemented rbmi: Methods based bootstrap standard error normal approximation percentile bootstrap methods. Denote treatment effect estimates \\(B\\) bootstrap samples \\(\\hat{\\theta}^*_b\\) (\\(b=1,\\ldots,B\\)). bootstrap standard error \\(\\hat{se}_{boot}\\) defined empirical standard deviation bootstrapped treatment effect estimates. Confidence intervals tests based bootstrap standard error can constructed way jackknife. Confidence intervals using percentile bootstrap based empirical quantiles bootstrap distribution corresponding statistical tests implemented rbmi via inversion confidence interval. Explicit formulas bootstrap inference implemented rbmi package considerations regarding required number bootstrap samples included Appendix Wolbers et al. (2022). simulation study reported Wolbers et al. (2022) demonstrated small inflation type error rate inference based bootstrap standard error (\\(5.3\\%\\) nominal type error rate \\(5\\%\\)) sample size n = 100 per group substantial amount missing data (>25% subjects ICE). Based simulations, recommend jackknife bootstrap inference performed better simulation study typically much faster compute bootstrap.","code":""},{"path":"/articles/stat_specs.html","id":"sec:poolbmlmi","dir":"Articles","previous_headings":"3 Statistical methodology","what":"Pooling step for inference of the bootstrapped MI methods","title":"rbmi: Statistical Specifications","text":"Assume analysis model applied \\(B\\times D\\) multiple imputed random datasets resulted \\(B\\times D\\) treatment effect estimates \\(\\hat{\\theta}_{bd}\\) (\\(b=1,\\ldots,B\\); \\(d=1,\\ldots,D\\)). final estimate treatment effect calculated sample mean \\(B*D\\) treatment effect estimates: \\[ \\hat{\\theta} = \\frac{1}{BD} \\sum_{b = 1}^B \\sum_{d = 1}^D \\hat{\\theta}_{bd}. \\] pooled variance based two components reflect variability within imputed bootstrap samples (von Hippel Bartlett (2021), formula 8.4): \\[ V(\\hat{\\theta}) = (1 + \\frac{1}{B})\\frac{MSB - MSW}{D} + \\frac{MSW}{BD} \\] \\(MSB\\) mean square bootstrapped datasets, \\(MSW\\) mean square within bootstrapped datasets imputed datasets: \\[ \\begin{align*} MSB &= \\frac{D}{B-1} \\sum_{b = 1}^B (\\bar{\\theta_{b}} - \\hat{\\theta})^2 \\\\ MSW &= \\frac{1}{B(D-1)} \\sum_{b = 1}^B \\sum_{d = 1}^D (\\theta_{bd} - \\bar{\\theta_b})^2 \\end{align*} \\] \\(\\bar{\\theta_{b}}\\) mean across \\(D\\) estimates obtained random imputation \\(b\\)-th bootstrap sample. degrees freedom estimated following formula (von Hippel Bartlett (2021), formula 8.6): \\[ \\nu = \\frac{(MSB\\cdot (B+1) - MSW\\cdot B)^2}{\\frac{MSB^2\\cdot (B+1)^2}{B-1} + \\frac{MSW^2\\cdot B}{D-1}} \\] Confidence intervals tests null hypothesis \\(H_0: \\theta=\\theta_0\\) based \\(t\\)-statistics \\(T\\): \\[ T= (\\hat{\\theta}-\\theta_0)/\\sqrt{V(\\hat{\\theta})}. \\] null hypothesis, \\(T\\) approximate \\(t\\)-distribution \\(\\nu\\) degrees freedom.","code":""},{"path":[]},{"path":"/articles/stat_specs.html","id":"treatment-effect-estimation","dir":"Articles","previous_headings":"3 Statistical methodology > 3.10 Comparison between the implemented approaches","what":"Treatment effect estimation","title":"rbmi: Statistical Specifications","text":"approaches provide consistent treatment effect estimates standard reference-based imputation methods case analysis model completed datasets general linear model ANCOVA. Methods conditional mean imputation also valid analysis models. validity conditional mean imputation formally demonstrated analyses using general linear model (Wolbers et al. (2022, sec. 2.4)) though may also applicable widely (e.g. general MMRM analysis models). Treatment effects based conditional mean imputation deterministic. methods affected Monte Carlo sampling error precision estimates depends number imputations bootstrap samples, respectively.","code":""},{"path":"/articles/stat_specs.html","id":"standard-errors-of-the-treatment-effect","dir":"Articles","previous_headings":"3 Statistical methodology > 3.10 Comparison between the implemented approaches","what":"Standard errors of the treatment effect","title":"rbmi: Statistical Specifications","text":"approaches imputation MAR assumption provide consistent estimates frequentist standard error. reference-based imputation methods, situation complicated two different types variance estimators proposed statistical literature (Bartlett (2023)). first frequentist variance describes actual repeated sampling variability estimator. reference-based missing data assumption correctly specified, resulting inference based variance correct frequentist sense, .e. hypothesis tests asymptotically correct type error control confidence intervals correct coverage probabilities repeated sampling (Bartlett (2023), Wolbers et al. (2022)). Reference-based missing data assumptions strong borrow information reference arm imputation active arm. consequence, size frequentist standard errors treatment effects may decrease increasing amounts missing data. second proposal -called “information-anchored” variance originally proposed context sensitivity analyses (Cro, Carpenter, Kenward (2019)). variance estimator based disentangling point estimation variance estimation altogether. information-anchoring principle described Cro, Carpenter, Kenward (2019) states relative increase variance treatment effect estimator MAR imputation increasing amounts missing data preserved reference-based imputation methods. resulting information-anchored variance typically similar variance MAR imputation typically increases increasing amounts missing data. However, information-anchored variance reflect actual variability reference-based estimator repeated sampling resulting inference highly conservative resulting substantial power loss (Wolbers et al. (2022)). Moreover, date, Bayesian frequentist framework developed information-anchored variance provides correct inference reference-based missingness assumptions, clear whether framework can even developed. Reference-based conditional mean imputation (method_condmean()) bootstrapped likelihood-based multiple methods (method = method_bmlmi()) obtain standard errors via resampling hence target frequentist variance (Wolbers et al. (2022), von Hippel Bartlett (2021)). finite samples, simulations sample size \\(n=100\\) per group reported Wolbers et al. (2022) demonstrated conditional mean imputation combined jackknife (method_condmean(type = \"jackknife\")) provided exact protection type one error rate whereas bootstrap (method_condmean(type = \"bootstrap\")) associated small type error inflation (5.1% 5.3% nominal level 5%). reference-based conditional mean imputation, alternative information-anchored variance can obtained following proposal Lu (2021). basic idea Lu (2021) obtain information-anchored variance via MAR imputation combined delta-adjustment delta selected data-driven way match reference-based estimator. conditional mean imputation, proposal Lu (2021) can implemented choosing delta-adjustment difference conditional mean imputation chosen reference-based assumption MAR original dataset. illustration different variances can obtained conditional mean imputation rbmi provided vignette “Frequentist information-anchored inference reference-based conditional mean imputation” (vignette(topic = \"CondMean_Inference\", package = \"rbmi\")). Reference-based Bayesian (approximate Bayesian) multiple imputation methods combined Rubin’s rules (method_bayes() method_approxbayes()) target information-anchored variance (Cro, Carpenter, Kenward (2019)). frequentist variance methods principle obtained via bootstrap jackknife re-sampling treatment effect estimates computationally intensive directly supported rbmi. view primary analyses, accurate type error control (can obtained using frequentist variance) important adherence information anchoring principle , us, fully compatible strong reference-based assumptions. case, reference-based imputation used primary analysis, critical chosen reference-based assumption can clinically justified, suitable sensitivity analyses conducted stress-test assumptions. Conditional mean imputation combined jackknife method leads deterministic standard error estimates , consequently, confidence intervals \\(p\\)-values also deterministic. particularly important regulatory setting important ascertain whether calculated \\(p\\)-value close critical boundary 5% truly threshold rather uncertain Monte Carlo error.","code":""},{"path":"/articles/stat_specs.html","id":"computational-complexity","dir":"Articles","previous_headings":"3 Statistical methodology > 3.10 Comparison between the implemented approaches","what":"Computational complexity","title":"rbmi: Statistical Specifications","text":"Bayesian MI methods rely specification prior distributions usage Markov chain Monte Carlo (MCMC) methods. methods based multiple imputation bootstrapping require tuning parameters specification number imputations \\(M\\) bootstrap samples \\(B\\) rely numerical optimization fitting MMRM imputation models via REML. Conditional mean imputation combined jackknife tuning parameters. rbmi implementation, fitting MMRM imputation model via REML computationally expensive. MCMC sampling using rstan (Stan Development Team (2020)) typically relatively fast setting requires small burn-burn-chains. addition, number random imputations reliable inference using Rubin’s rules often smaller number resamples required jackknife bootstrap (see e.g. discussions . R. White, Royston, Wood (2011, sec. 7) Bayesian MI Appendix Wolbers et al. (2022) bootstrap). Thus, many applications, expect conventional MI based Bayesian posterior draws fastest, followed conventional MI using approximate Bayesian posterior draws conditional mean imputation combined jackknife. Conditional mean imputation combined bootstrap bootstrapped MI methods typically computationally demanding. note, implemented methods conceptually straightforward parallelise parallelisation support provided rbmi.","code":""},{"path":"/articles/stat_specs.html","id":"sec:rbmiFunctions","dir":"Articles","previous_headings":"","what":"Mapping of statistical methods to rbmi functions","title":"rbmi: Statistical Specifications","text":"full documentation rbmi package functionality refer help pages functions package vignettes. give brief overview different steps imputation procedure mapped rbmi functions: Bayesian posterior parameter draws imputation model obtained via argument method = method_bayes(). Approximate Bayesian posterior parameter draws imputation model obtained via argument method = method_approxbayes(). ML REML parameter estimates imputation model parameters original dataset leave-one-subject-datasets (required jackknife) obtained via argument method = method_condmean(type = \"jackknife\"). ML REML parameter estimates imputation model parameters original dataset bootstrapped datasets obtained via argument method = method_condmean(type = \"bootstrap\"). Bootstrapped MI methods obtained via argument method = method_bmlmi(B=B, D=D) \\(B\\) refers number bootstrap samples \\(D\\) number random imputations bootstrap sample. imputation step using random imputation deterministic conditional mean imputation, respectively, implemented function impute(). Imputation can performed assuming already implemented imputation strategies presented section 3.4. Additionally, user-defined imputation strategies also supported. analysis step implemented function analyse() applies analysis model imputed datasets. default, analysis model (argument fun) ancova() function alternative analysis functions can also provided user. analyse() function also allows \\(\\delta\\)-adjustments imputed datasets prior analysis via argument delta. inference step implemented function pool() pools results across imputed datasets. Rubin Bernard rule applied case (approximate) Bayesian MI. conditional mean imputation, jackknife bootstrap (normal approximation percentile) inference supported. BMLMI, pooling inference steps performed via pool() case implements method described Section 3.9.","code":""},{"path":"/articles/stat_specs.html","id":"sec:otherSoftware","dir":"Articles","previous_headings":"","what":"Comparison to other software implementations","title":"rbmi: Statistical Specifications","text":"established software implementation reference-based imputation SAS -called “five macros” James Roger (Roger (2021)). alternative R implementation also currently development R package RefBasedMI (McGrath White (2021)). rbmi several features supported implementations: addition Bayesian MI approach implemented also packages, implementation provides three alternative MI approaches: approximate Bayesian MI, conditional mean imputation combined resampling, bootstrapped MI. rbmi allows usage data collected ICE. example, suppose want adopt treatment policy strategy ICE “treatment discontinuation”. possible implementation strategy use observed outcome data subjects remain study ICE use reference-based imputation case subject drops . implementation, implemented excluding observed post ICE data imputation model assumes MAR missingness including analysis model. knowledge, directly supported implementations. RefBasedMI fits imputation model data treatment group separately implies covariate-treatment group interactions covariates pooled data treatment groups. contrast, Roger’s five macros assume joint model including data randomized groups covariate-treatment interactions covariates allowed. also chose implement joint model use flexible model linear predictor may may include interaction term covariate treatment group. addition, imputation model also allows inclusion time-varying covariates. implementation, grouping subjects purpose imputation model (definition reference group) need correspond assigned treatment groups. provides additional flexibility imputation procedure. clear us whether feature supported Roger’s five macros RefBasedMI. believe R-based implementation modular RefBasedMI facilitate package enhancements. contrast, general causal model introduced . White, Royes, Best (2020) available implementations currently supported .","code":""},{"path":[]},{"path":"/authors.html","id":null,"dir":"","previous_headings":"","what":"Authors","title":"Authors and Citation","text":"Craig Gower-Page. Author, maintainer. Alessandro Noci. Author. Marcel Wolbers. Contributor. Roche. Copyright holder, funder.","code":""},{"path":"/authors.html","id":"citation","dir":"","previous_headings":"","what":"Citation","title":"Authors and Citation","text":"Gower-Page C, Noci (2024). rbmi: Reference Based Multiple Imputation. R package version 1.2.6, https://github.com/insightsengineering/rbmi, https://insightsengineering.github.io/rbmi/main/.","code":"@Manual{, title = {rbmi: Reference Based Multiple Imputation}, author = {Craig Gower-Page and Alessandro Noci}, year = {2024}, note = {R package version 1.2.6, https://github.com/insightsengineering/rbmi}, url = {https://insightsengineering.github.io/rbmi/main/}, }"},{"path":[]},{"path":"/index.html","id":"overview","dir":"","previous_headings":"","what":"Overview","title":"Reference Based Multiple Imputation","text":"rbmi R package imputation missing data clinical trials continuous multivariate normal longitudinal outcomes. supports imputation missing random (MAR) assumption, reference-based imputation methods, delta adjustments (required sensitivity analysis tipping point analyses). package implements Bayesian approximate Bayesian multiple imputation combined Rubin’s rules inference, frequentist conditional mean imputation combined (jackknife bootstrap) resampling.","code":""},{"path":"/index.html","id":"installation","dir":"","previous_headings":"","what":"Installation","title":"Reference Based Multiple Imputation","text":"package can installed directly CRAN via:","code":"install.packages(\"rbmi\")"},{"path":"/index.html","id":"usage","dir":"","previous_headings":"","what":"Usage","title":"Reference Based Multiple Imputation","text":"package designed around 4 core functions: draws() - Fits multiple imputation models impute() - Imputes multiple datasets analyse() - Analyses multiple datasets pool() - Pools multiple results single statistic basic usage core functions described quickstart vignette:","code":"vignette(topic = \"quickstart\", package = \"rbmi\")"},{"path":"/index.html","id":"support","dir":"","previous_headings":"","what":"Support","title":"Reference Based Multiple Imputation","text":"help regards using package find bug please create GitHub issue","code":""},{"path":"/reference/QR_decomp.html","id":null,"dir":"Reference","previous_headings":"","what":"QR decomposition — QR_decomp","title":"QR decomposition — QR_decomp","text":"QR decomposition defined Stan user's guide (section 1.2).","code":""},{"path":"/reference/QR_decomp.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"QR decomposition — QR_decomp","text":"","code":"QR_decomp(mat)"},{"path":"/reference/QR_decomp.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"QR decomposition — QR_decomp","text":"mat matrix perform QR decomposition .","code":""},{"path":"/reference/Stack.html","id":null,"dir":"Reference","previous_headings":"","what":"R6 Class for a FIFO stack — Stack","title":"R6 Class for a FIFO stack — Stack","text":"simple stack object offering add / pop functionality","code":""},{"path":"/reference/Stack.html","id":"public-fields","dir":"Reference","previous_headings":"","what":"Public fields","title":"R6 Class for a FIFO stack — Stack","text":"stack list containing current stack","code":""},{"path":[]},{"path":"/reference/Stack.html","id":"public-methods","dir":"Reference","previous_headings":"","what":"Public methods","title":"R6 Class for a FIFO stack — Stack","text":"Stack$add() Stack$pop() Stack$clone()","code":""},{"path":"/reference/Stack.html","id":"method-add-","dir":"Reference","previous_headings":"","what":"Method add()","title":"R6 Class for a FIFO stack — Stack","text":"Adds content end stack (must list)","code":""},{"path":"/reference/Stack.html","id":"usage","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for a FIFO stack — Stack","text":"","code":"Stack$add(x)"},{"path":"/reference/Stack.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for a FIFO stack — Stack","text":"x content add stack","code":""},{"path":"/reference/Stack.html","id":"method-pop-","dir":"Reference","previous_headings":"","what":"Method pop()","title":"R6 Class for a FIFO stack — Stack","text":"Retrieve content stack","code":""},{"path":"/reference/Stack.html","id":"usage-1","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for a FIFO stack — Stack","text":"","code":"Stack$pop(i)"},{"path":"/reference/Stack.html","id":"arguments-1","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for a FIFO stack — Stack","text":"number items retrieve stack. less items left stack just return everything left.","code":""},{"path":"/reference/Stack.html","id":"method-clone-","dir":"Reference","previous_headings":"","what":"Method clone()","title":"R6 Class for a FIFO stack — Stack","text":"objects class cloneable method.","code":""},{"path":"/reference/Stack.html","id":"usage-2","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for a FIFO stack — Stack","text":"","code":"Stack$clone(deep = FALSE)"},{"path":"/reference/Stack.html","id":"arguments-2","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for a FIFO stack — Stack","text":"deep Whether make deep clone.","code":""},{"path":"/reference/add_class.html","id":null,"dir":"Reference","previous_headings":"","what":"Add a class — add_class","title":"Add a class — add_class","text":"Utility function add class object. Adds new class existing classes.","code":""},{"path":"/reference/add_class.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Add a class — add_class","text":"","code":"add_class(x, cls)"},{"path":"/reference/add_class.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Add a class — add_class","text":"x object add class . cls class added.","code":""},{"path":"/reference/adjust_trajectories.html","id":null,"dir":"Reference","previous_headings":"","what":"Adjust trajectories due to the intercurrent event (ICE) — adjust_trajectories","title":"Adjust trajectories due to the intercurrent event (ICE) — adjust_trajectories","text":"Adjust trajectories due intercurrent event (ICE)","code":""},{"path":"/reference/adjust_trajectories.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Adjust trajectories due to the intercurrent event (ICE) — adjust_trajectories","text":"","code":"adjust_trajectories( distr_pars_group, outcome, ids, ind_ice, strategy_fun, distr_pars_ref = NULL )"},{"path":"/reference/adjust_trajectories.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Adjust trajectories due to the intercurrent event (ICE) — adjust_trajectories","text":"distr_pars_group Named list containing simulation parameters multivariate normal distribution assumed given treatment group. contains following elements: mu: Numeric vector indicating mean outcome trajectory. include outcome baseline. sigma Covariance matrix outcome trajectory. outcome Numeric variable specifies longitudinal outcome. ids Factor variable specifies id subject. ind_ice binary variable takes value 1 corresponding outcome affected ICE 0 otherwise. strategy_fun Function implementing trajectories intercurrent event (ICE). Must one getStrategies(). See getStrategies() details. distr_pars_ref Optional. Named list containing simulation parameters reference arm. contains following elements: mu: Numeric vector indicating mean outcome trajectory assuming ICEs. include outcome baseline. sigma Covariance matrix outcome trajectory assuming ICEs.","code":""},{"path":"/reference/adjust_trajectories.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Adjust trajectories due to the intercurrent event (ICE) — adjust_trajectories","text":"numeric vector containing adjusted trajectories.","code":""},{"path":[]},{"path":"/reference/adjust_trajectories_single.html","id":null,"dir":"Reference","previous_headings":"","what":"Adjust trajectory of a subject's outcome due to the intercurrent event (ICE) — adjust_trajectories_single","title":"Adjust trajectory of a subject's outcome due to the intercurrent event (ICE) — adjust_trajectories_single","text":"Adjust trajectory subject's outcome due intercurrent event (ICE)","code":""},{"path":"/reference/adjust_trajectories_single.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Adjust trajectory of a subject's outcome due to the intercurrent event (ICE) — adjust_trajectories_single","text":"","code":"adjust_trajectories_single( distr_pars_group, outcome, strategy_fun, distr_pars_ref = NULL )"},{"path":"/reference/adjust_trajectories_single.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Adjust trajectory of a subject's outcome due to the intercurrent event (ICE) — adjust_trajectories_single","text":"distr_pars_group Named list containing simulation parameters multivariate normal distribution assumed given treatment group. contains following elements: mu: Numeric vector indicating mean outcome trajectory. include outcome baseline. sigma Covariance matrix outcome trajectory. outcome Numeric variable specifies longitudinal outcome. strategy_fun Function implementing trajectories intercurrent event (ICE). Must one getStrategies(). See getStrategies() details. distr_pars_ref Optional. Named list containing simulation parameters reference arm. contains following elements: mu: Numeric vector indicating mean outcome trajectory assuming ICEs. include outcome baseline. sigma Covariance matrix outcome trajectory assuming ICEs.","code":""},{"path":"/reference/adjust_trajectories_single.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Adjust trajectory of a subject's outcome due to the intercurrent event (ICE) — adjust_trajectories_single","text":"numeric vector containing adjusted trajectory single subject.","code":""},{"path":"/reference/adjust_trajectories_single.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Adjust trajectory of a subject's outcome due to the intercurrent event (ICE) — adjust_trajectories_single","text":"outcome specified --post-ICE observations (.e. observations adjusted) set NA.","code":""},{"path":"/reference/analyse.html","id":null,"dir":"Reference","previous_headings":"","what":"Analyse Multiple Imputed Datasets — analyse","title":"Analyse Multiple Imputed Datasets — analyse","text":"function takes multiple imputed datasets (generated impute() function) runs analysis function .","code":""},{"path":"/reference/analyse.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Analyse Multiple Imputed Datasets — analyse","text":"","code":"analyse(imputations, fun = ancova, delta = NULL, ...)"},{"path":"/reference/analyse.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Analyse Multiple Imputed Datasets — analyse","text":"imputations imputations object created impute(). fun analysis function applied imputed dataset. See details. delta data.frame containing delta transformation applied imputed datasets prior running fun. See details. ... Additional arguments passed onto fun.","code":""},{"path":"/reference/analyse.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Analyse Multiple Imputed Datasets — analyse","text":"function works performing following steps: Extract dataset imputations object. Apply delta adjustments specified delta argument. Run analysis function fun dataset. Repeat steps 1-3 across datasets inside imputations object. Collect return analysis results. analysis function fun must take data.frame first argument. options analyse() passed onto fun via .... fun must return named list element list containing single numeric element called est (additionally se df originally specified method_bayes() method_approxbayes()) .e.: Please note vars$subjid column (defined original call draws()) scrambled data.frames provided fun. say contain original subject values hard coding subject ids strictly avoided. default fun ancova() function. Please note function requires vars object, created set_vars(), provided via vars argument e.g. analyse(imputeObj, vars = set_vars(...)). Please see documentation ancova() full details. Please also note theoretical justification conditional mean imputation method (method = method_condmean() draws()) relies fact ANCOVA linear transformation outcomes. Thus care required applying alternative analysis functions setting. delta argument can used specify offsets applied outcome variable imputed datasets prior analysis. typically used sensitivity tipping point analyses. delta dataset must contain columns vars$subjid, vars$visit (specified original call draws()) delta. Essentially data.frame merged onto imputed dataset vars$subjid vars$visit outcome variable modified : Please note order provide maximum flexibility, delta argument can used modify /outcome values including imputed. Care must taken defining offsets. recommend use helper function delta_template() define delta datasets provides utility variables is_missing can used identify exactly visits imputed.","code":"myfun <- function(dat, ...) { mod_1 <- lm(data = dat, outcome ~ group) mod_2 <- lm(data = dat, outcome ~ group + covar) x <- list( trt_1 = list( est = coef(mod_1)[[group]], se = sqrt(vcov(mod_1)[group, group]), df = df.residual(mod_1) ), trt_2 = list( est = coef(mod_2)[[group]], se = sqrt(vcov(mod_2)[group, group]), df = df.residual(mod_2) ) ) return(x) } imputed_data[[vars$outcome]] <- imputed_data[[vars$outcome]] + imputed_data[[\"delta\"]]"},{"path":[]},{"path":"/reference/analyse.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Analyse Multiple Imputed Datasets — analyse","text":"","code":"if (FALSE) { vars <- set_vars( subjid = \"subjid\", visit = \"visit\", outcome = \"outcome\", group = \"group\", covariates = c(\"sex\", \"age\", \"sex*age\") ) analyse( imputations = imputeObj, vars = vars ) deltadf <- data.frame( subjid = c(\"Pt1\", \"Pt1\", \"Pt2\"), visit = c(\"Visit_1\", \"Visit_2\", \"Visit_2\"), delta = c( 5, 9, -10) ) analyse( imputations = imputeObj, delta = deltadf, vars = vars ) }"},{"path":"/reference/ancova.html","id":null,"dir":"Reference","previous_headings":"","what":"Analysis of Covariance — ancova","title":"Analysis of Covariance — ancova","text":"Performs analysis covariance two groups returning estimated \"treatment effect\" (.e. contrast two treatment groups) least square means estimates group.","code":""},{"path":"/reference/ancova.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Analysis of Covariance — ancova","text":"","code":"ancova(data, vars, visits = NULL, weights = c(\"proportional\", \"equal\"))"},{"path":"/reference/ancova.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Analysis of Covariance — ancova","text":"data data.frame containing data used model. vars vars object generated set_vars(). group, visit, outcome covariates elements required. See details. visits optional character vector specifying visits fit ancova model . NULL, separate ancova model fit outcomes visit (determined unique(data[[vars$visit]])). See details. weights Character, either \"proportional\" (default) \"equal\". Specifies weighting strategy used categorical covariates calculating lsmeans. See details.","code":""},{"path":"/reference/ancova.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Analysis of Covariance — ancova","text":"function works follows: Select first value visits. Subset data observations occurred visit. Fit linear model vars$outcome ~ vars$group + vars$covariates. Extract \"treatment effect\" & least square means treatment group. Repeat points 2-3 values visits. value visits provided set unique(data[[vars$visit]]). order meet formatting standards set analyse() results collapsed single list suffixed visit name, e.g.: Please note \"ref\" refers first factor level vars$group necessarily coincide control arm. Analogously, \"alt\" refers second factor level vars$group. \"trt\" refers model contrast translating mean difference second level first level. want include interaction terms model can done providing covariates argument set_vars() e.g. set_vars(covariates = c(\"sex*age\")).","code":"list( trt_visit_1 = list(est = ...), lsm_ref_visit_1 = list(est = ...), lsm_alt_visit_1 = list(est = ...), trt_visit_2 = list(est = ...), lsm_ref_visit_2 = list(est = ...), lsm_alt_visit_2 = list(est = ...), ... )"},{"path":"/reference/ancova.html","id":"weighting","dir":"Reference","previous_headings":"","what":"Weighting","title":"Analysis of Covariance — ancova","text":"\"proportional\" default scheme used. equivalent standardization, .e. lsmeans group equal predicted mean outcome ancova model group based baseline characteristics subjects regardless assigned group. alternative weighting scheme, \"equal\", creates hypothetical patients expanding combinations models categorical covariates. lsmeans calculated average predicted mean outcome hypothetical patients assuming come group turn. short: \"proportional\" weights categorical covariates based upon frequency occurrence data. \"equal\" weights categorical covariates equally across theoretical combinations.","code":""},{"path":[]},{"path":"/reference/ancova_single.html","id":null,"dir":"Reference","previous_headings":"","what":"Implements an Analysis of Covariance (ANCOVA) — ancova_single","title":"Implements an Analysis of Covariance (ANCOVA) — ancova_single","text":"Performance analysis covariance. See ancova() full details.","code":""},{"path":"/reference/ancova_single.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Implements an Analysis of Covariance (ANCOVA) — ancova_single","text":"","code":"ancova_single( data, outcome, group, covariates, weights = c(\"proportional\", \"equal\") )"},{"path":"/reference/ancova_single.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Implements an Analysis of Covariance (ANCOVA) — ancova_single","text":"data data.frame containing data required model. outcome Character, name outcome variable data. group Character, name group variable data. covariates Character vector containing name additional covariates included model well interaction terms. weights Character, specifies whether use \"proportional\" \"equal\" weighting categorical covariate combination calculating lsmeans.","code":""},{"path":"/reference/ancova_single.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Implements an Analysis of Covariance (ANCOVA) — ancova_single","text":"group must factor variable 2 levels. outcome must continuous numeric variable.","code":""},{"path":[]},{"path":"/reference/ancova_single.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Implements an Analysis of Covariance (ANCOVA) — ancova_single","text":"","code":"if (FALSE) { iris2 <- iris[ iris$Species %in% c(\"versicolor\", \"virginica\"), ] iris2$Species <- factor(iris2$Species) ancova_single(iris2, \"Sepal.Length\", \"Species\", c(\"Petal.Length * Petal.Width\")) }"},{"path":"/reference/antidepressant_data.html","id":null,"dir":"Reference","previous_headings":"","what":"Antidepressant trial data — antidepressant_data","title":"Antidepressant trial data — antidepressant_data","text":"dataset containing data publicly available example data set antidepressant clinical trial. dataset available website Drug Information Association Scientific Working Group Estimands Missing Data. per website, original data antidepressant clinical trial four treatments; two doses experimental medication, positive control, placebo published Goldstein et al (2004). mask real data, week 8 observations removed two arms created: original placebo arm \"drug arm\" created randomly selecting patients three non-placebo arms.","code":""},{"path":"/reference/antidepressant_data.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Antidepressant trial data — antidepressant_data","text":"","code":"antidepressant_data"},{"path":"/reference/antidepressant_data.html","id":"format","dir":"Reference","previous_headings":"","what":"Format","title":"Antidepressant trial data — antidepressant_data","text":"data.frame 608 rows 11 variables: PATIENT: patients IDs. HAMATOTL: total score Hamilton Anxiety Rating Scale. PGIIMP: patient's Global Impression Improvement Rating Scale. RELDAYS: number days visit baseline. VISIT: post-baseline visit. levels 4,5,6,7. THERAPY: treatment group variable. equal PLACEBO observations placebo arm, DRUG observations active arm. GENDER: patient's gender. POOLINV: pooled investigator. BASVAL: baseline outcome value. HAMDTL17: Hamilton 17-item rating scale value. CHANGE: change baseline Hamilton 17-item rating scale.","code":""},{"path":"/reference/antidepressant_data.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Antidepressant trial data — antidepressant_data","text":"relevant endpoint Hamilton 17-item rating scale depression (HAMD17) baseline weeks 1, 2, 4, 6 assessments included. Study drug discontinuation occurred 24% subjects active drug 26% placebo. data study drug discontinuation missing single additional intermittent missing observation.","code":""},{"path":"/reference/antidepressant_data.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Antidepressant trial data — antidepressant_data","text":"Goldstein, Lu, Detke, Wiltse, Mallinckrodt, Demitrack. Duloxetine treatment depression: double-blind placebo-controlled comparison paroxetine. J Clin Psychopharmacol 2004;24: 389-399.","code":""},{"path":"/reference/apply_delta.html","id":null,"dir":"Reference","previous_headings":"","what":"Applies delta adjustment — apply_delta","title":"Applies delta adjustment — apply_delta","text":"Takes delta dataset adjusts outcome variable adding corresponding delta.","code":""},{"path":"/reference/apply_delta.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Applies delta adjustment — apply_delta","text":"","code":"apply_delta(data, delta = NULL, group = NULL, outcome = NULL)"},{"path":"/reference/apply_delta.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Applies delta adjustment — apply_delta","text":"data data.frame outcome column adjusted. delta data.frame (must contain column called delta). group character vector variables data delta used merge 2 data.frames together . outcome character, name outcome variable data.","code":""},{"path":"/reference/as_analysis.html","id":null,"dir":"Reference","previous_headings":"","what":"Construct an analysis object — as_analysis","title":"Construct an analysis object — as_analysis","text":"Creates analysis object ensuring components correctly defined.","code":""},{"path":"/reference/as_analysis.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Construct an analysis object — as_analysis","text":"","code":"as_analysis(results, method, delta = NULL, fun = NULL, fun_name = NULL)"},{"path":"/reference/as_analysis.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Construct an analysis object — as_analysis","text":"results list lists contain analysis results imputation See analyse() details object look like. method method object specified draws(). delta delta dataset used. See analyse() details specified. fun analysis function used. fun_name character name analysis function (used printing) purposes.","code":""},{"path":"/reference/as_ascii_table.html","id":null,"dir":"Reference","previous_headings":"","what":"as_ascii_table — as_ascii_table","title":"as_ascii_table — as_ascii_table","text":"function takes data.frame attempts convert simple ascii format suitable printing screen assumed variable values .character() method order cast character.","code":""},{"path":"/reference/as_ascii_table.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"as_ascii_table — as_ascii_table","text":"","code":"as_ascii_table(dat, line_prefix = \" \", pcol = NULL)"},{"path":"/reference/as_ascii_table.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"as_ascii_table — as_ascii_table","text":"dat Input dataset convert ascii table line_prefix Symbols prefix infront every line table pcol name column handled p-value. Sets value <0.001 value 0 rounding","code":""},{"path":"/reference/as_class.html","id":null,"dir":"Reference","previous_headings":"","what":"Set Class — as_class","title":"Set Class — as_class","text":"Utility function set objects class.","code":""},{"path":"/reference/as_class.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Set Class — as_class","text":"","code":"as_class(x, cls)"},{"path":"/reference/as_class.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Set Class — as_class","text":"x object set class . cls class set.","code":""},{"path":"/reference/as_cropped_char.html","id":null,"dir":"Reference","previous_headings":"","what":"as_cropped_char — as_cropped_char","title":"as_cropped_char — as_cropped_char","text":"Makes character string x chars Reduce x char string ...","code":""},{"path":"/reference/as_cropped_char.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"as_cropped_char — as_cropped_char","text":"","code":"as_cropped_char(inval, crop_at = 30, ndp = 3)"},{"path":"/reference/as_cropped_char.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"as_cropped_char — as_cropped_char","text":"inval single element value crop_at character limit ndp Number decimal places display","code":""},{"path":"/reference/as_dataframe.html","id":null,"dir":"Reference","previous_headings":"","what":"Convert object to dataframe — as_dataframe","title":"Convert object to dataframe — as_dataframe","text":"Convert object dataframe","code":""},{"path":"/reference/as_dataframe.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Convert object to dataframe — as_dataframe","text":"","code":"as_dataframe(x)"},{"path":"/reference/as_dataframe.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Convert object to dataframe — as_dataframe","text":"x data.frame like object Utility function convert \"data.frame-like\" object actual data.frame avoid issues inconsistency methods ( [() dplyr's grouped dataframes)","code":""},{"path":"/reference/as_draws.html","id":null,"dir":"Reference","previous_headings":"","what":"Creates a draws object — as_draws","title":"Creates a draws object — as_draws","text":"Creates draws object final output call draws().","code":""},{"path":"/reference/as_draws.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Creates a draws object — as_draws","text":"","code":"as_draws(method, samples, data, formula, n_failures = NULL, fit = NULL)"},{"path":"/reference/as_draws.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Creates a draws object — as_draws","text":"method method object generated either method_bayes(), method_approxbayes(), method_condmean() method_bmlmi(). samples list sample_single objects. See sample_single(). data R6 longdata object containing relevant input data information. formula Fixed effects formula object used model specification. n_failures Absolute number failures model fit. fit method_bayes() chosen, returns MCMC Stan fit object. Otherwise NULL.","code":""},{"path":"/reference/as_draws.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Creates a draws object — as_draws","text":"draws object named list containing following: data: R6 longdata object containing relevant input data information. method: method object generated either method_bayes(), method_approxbayes() method_condmean(). samples: list containing estimated parameters interest. element samples named list containing following: ids: vector characters containing ids subjects included original dataset. beta: numeric vector estimated regression coefficients. sigma: list estimated covariance matrices (one level vars$group). theta: numeric vector transformed covariances. failed: Logical. TRUE model fit failed. ids_samp: vector characters containing ids subjects included given sample. fit: method_bayes() chosen, returns MCMC Stan fit object. Otherwise NULL. n_failures: absolute number failures model fit. Relevant method_condmean(type = \"bootstrap\"), method_approxbayes() method_bmlmi(). formula: fixed effects formula object used model specification.","code":""},{"path":"/reference/as_imputation.html","id":null,"dir":"Reference","previous_headings":"","what":"Create an imputation object — as_imputation","title":"Create an imputation object — as_imputation","text":"function creates object returned impute(). Essentially glorified wrapper around list() ensuring required elements set class added expected.","code":""},{"path":"/reference/as_imputation.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create an imputation object — as_imputation","text":"","code":"as_imputation(imputations, data, method, references)"},{"path":"/reference/as_imputation.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create an imputation object — as_imputation","text":"imputations list imputations_list's created imputation_df() data longdata object created longDataConstructor() method method object created method_condmean(), method_bayes() method_approxbayes() references named vector. Identifies references used generating imputed values. form c(\"Group\" = \"Reference\", \"Group\" = \"Reference\").","code":""},{"path":"/reference/as_indices.html","id":null,"dir":"Reference","previous_headings":"","what":"Convert indicator to index — as_indices","title":"Convert indicator to index — as_indices","text":"Converts string 0's 1's index positions 1's padding results 0's length","code":""},{"path":"/reference/as_indices.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Convert indicator to index — as_indices","text":"","code":"as_indices(x)"},{"path":"/reference/as_indices.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Convert indicator to index — as_indices","text":"x character vector whose values either \"0\" \"1\". elements vector must length","code":""},{"path":"/reference/as_indices.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Convert indicator to index — as_indices","text":".e.","code":"patmap(c(\"1101\", \"0001\")) -> list(c(1,2,4,999), c(4,999, 999, 999))"},{"path":"/reference/as_mmrm_df.html","id":null,"dir":"Reference","previous_headings":"","what":"Creates a ","title":"Creates a ","text":"Converts design matrix + key variables common format particular function following: Renames covariates V1, V2, etc avoid issues special characters variable names Ensures key variables right type Inserts outcome, visit subjid variables data.frame naming outcome, visit subjid provided also insert group variable data.frame named group","code":""},{"path":"/reference/as_mmrm_df.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Creates a ","text":"","code":"as_mmrm_df(designmat, outcome, visit, subjid, group = NULL)"},{"path":"/reference/as_mmrm_df.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Creates a ","text":"designmat data.frame matrix containing covariates use MMRM model. Dummy variables must already expanded , .e. via stats::model.matrix(). contain missing values outcome numeric vector. outcome value regressed MMRM model. visit character / factor vector. Indicates visit outcome value occurred . subjid character / factor vector. subject identifier used link separate visits belong subject. group character / factor vector. Indicates treatment group patient belongs .","code":""},{"path":"/reference/as_mmrm_formula.html","id":null,"dir":"Reference","previous_headings":"","what":"Create MMRM formula — as_mmrm_formula","title":"Create MMRM formula — as_mmrm_formula","text":"Derives MMRM model formula structure mmrm_df. returns formula object form:","code":""},{"path":"/reference/as_mmrm_formula.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create MMRM formula — as_mmrm_formula","text":"","code":"as_mmrm_formula(mmrm_df, cov_struct)"},{"path":"/reference/as_mmrm_formula.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create MMRM formula — as_mmrm_formula","text":"mmrm_df mmrm data.frame created as_mmrm_df() cov_struct Character - covariance structure used, must one \"us\", \"toep\", \"cs\", \"ar1\"","code":""},{"path":"/reference/as_mmrm_formula.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Create MMRM formula — as_mmrm_formula","text":"","code":"outcome ~ 0 + V1 + V2 + V4 + ... + us(visit | group / subjid)"},{"path":"/reference/as_model_df.html","id":null,"dir":"Reference","previous_headings":"","what":"Expand data.frame into a design matrix — as_model_df","title":"Expand data.frame into a design matrix — as_model_df","text":"Expands data.frame using formula create design matrix. Key details always place outcome variable first column return object.","code":""},{"path":"/reference/as_model_df.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Expand data.frame into a design matrix — as_model_df","text":"","code":"as_model_df(dat, frm)"},{"path":"/reference/as_model_df.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Expand data.frame into a design matrix — as_model_df","text":"dat data.frame frm formula","code":""},{"path":"/reference/as_model_df.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Expand data.frame into a design matrix — as_model_df","text":"outcome column may contain NA's none variables listed formula contain missing values","code":""},{"path":"/reference/as_simple_formula.html","id":null,"dir":"Reference","previous_headings":"","what":"Creates a simple formula object from a string — as_simple_formula","title":"Creates a simple formula object from a string — as_simple_formula","text":"Converts string list variables formula object","code":""},{"path":"/reference/as_simple_formula.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Creates a simple formula object from a string — as_simple_formula","text":"","code":"as_simple_formula(outcome, covars)"},{"path":"/reference/as_simple_formula.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Creates a simple formula object from a string — as_simple_formula","text":"outcome character (length 1 vector). Name outcome variable covars character (vector). Name covariates","code":""},{"path":"/reference/as_simple_formula.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Creates a simple formula object from a string — as_simple_formula","text":"formula","code":""},{"path":"/reference/as_stan_array.html","id":null,"dir":"Reference","previous_headings":"","what":"As array — as_stan_array","title":"As array — as_stan_array","text":"Converts numeric value length 1 1 dimension array. avoid type errors thrown stan length 1 numeric vectors provided R stan::vector inputs","code":""},{"path":"/reference/as_stan_array.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"As array — as_stan_array","text":"","code":"as_stan_array(x)"},{"path":"/reference/as_stan_array.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"As array — as_stan_array","text":"x numeric vector","code":""},{"path":"/reference/as_strata.html","id":null,"dir":"Reference","previous_headings":"","what":"Create vector of Stratas — as_strata","title":"Create vector of Stratas — as_strata","text":"Collapse multiple categorical variables distinct unique categories. e.g. return","code":"as_strata(c(1,1,2,2,2,1), c(5,6,5,5,6,5)) c(1,2,3,3,4,1)"},{"path":"/reference/as_strata.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create vector of Stratas — as_strata","text":"","code":"as_strata(...)"},{"path":"/reference/as_strata.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create vector of Stratas — as_strata","text":"... numeric/character/factor vectors length","code":""},{"path":"/reference/as_strata.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Create vector of Stratas — as_strata","text":"","code":"if (FALSE) { as_strata(c(1,1,2,2,2,1), c(5,6,5,5,6,5)) }"},{"path":"/reference/assert_variables_exist.html","id":null,"dir":"Reference","previous_headings":"","what":"Assert that all variables exist within a dataset — assert_variables_exist","title":"Assert that all variables exist within a dataset — assert_variables_exist","text":"Performs assertion check ensure vector variable exists within data.frame expected.","code":""},{"path":"/reference/assert_variables_exist.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Assert that all variables exist within a dataset — assert_variables_exist","text":"","code":"assert_variables_exist(data, vars)"},{"path":"/reference/assert_variables_exist.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Assert that all variables exist within a dataset — assert_variables_exist","text":"data data.frame vars character vector variable names","code":""},{"path":"/reference/char2fct.html","id":null,"dir":"Reference","previous_headings":"","what":"Convert character variables to factor — char2fct","title":"Convert character variables to factor — char2fct","text":"Provided vector variable names function converts character variables factors. affect numeric existing factor variables","code":""},{"path":"/reference/char2fct.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Convert character variables to factor — char2fct","text":"","code":"char2fct(data, vars = NULL)"},{"path":"/reference/char2fct.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Convert character variables to factor — char2fct","text":"data data.frame vars character vector variables data","code":""},{"path":"/reference/check_ESS.html","id":null,"dir":"Reference","previous_headings":"","what":"Diagnostics of the MCMC based on ESS — check_ESS","title":"Diagnostics of the MCMC based on ESS — check_ESS","text":"Check quality MCMC draws posterior distribution checking whether relative ESS sufficiently large.","code":""},{"path":"/reference/check_ESS.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Diagnostics of the MCMC based on ESS — check_ESS","text":"","code":"check_ESS(stan_fit, n_draws, threshold_lowESS = 0.4)"},{"path":"/reference/check_ESS.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Diagnostics of the MCMC based on ESS — check_ESS","text":"stan_fit stanfit object. n_draws Number MCMC draws. threshold_lowESS number [0,1] indicating minimum acceptable value relative ESS. See details.","code":""},{"path":"/reference/check_ESS.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Diagnostics of the MCMC based on ESS — check_ESS","text":"warning message case detected problems.","code":""},{"path":"/reference/check_ESS.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Diagnostics of the MCMC based on ESS — check_ESS","text":"check_ESS() works follows: Extract ESS stan_fit parameter model. Compute relative ESS (.e. ESS divided number draws). Check whether parameter ESS lower threshold. least one parameter relative ESS threshold, warning thrown.","code":""},{"path":"/reference/check_hmc_diagn.html","id":null,"dir":"Reference","previous_headings":"","what":"Diagnostics of the MCMC based on HMC-related measures. — check_hmc_diagn","title":"Diagnostics of the MCMC based on HMC-related measures. — check_hmc_diagn","text":"Check : divergent iterations. Bayesian Fraction Missing Information (BFMI) sufficiently low. number iterations saturated max treedepth zero. Please see rstan::check_hmc_diagnostics() details.","code":""},{"path":"/reference/check_hmc_diagn.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Diagnostics of the MCMC based on HMC-related measures. — check_hmc_diagn","text":"","code":"check_hmc_diagn(stan_fit)"},{"path":"/reference/check_hmc_diagn.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Diagnostics of the MCMC based on HMC-related measures. — check_hmc_diagn","text":"stan_fit stanfit object.","code":""},{"path":"/reference/check_hmc_diagn.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Diagnostics of the MCMC based on HMC-related measures. — check_hmc_diagn","text":"warning message case detected problems.","code":""},{"path":"/reference/check_mcmc.html","id":null,"dir":"Reference","previous_headings":"","what":"Diagnostics of the MCMC — check_mcmc","title":"Diagnostics of the MCMC — check_mcmc","text":"Diagnostics MCMC","code":""},{"path":"/reference/check_mcmc.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Diagnostics of the MCMC — check_mcmc","text":"","code":"check_mcmc(stan_fit, n_draws, threshold_lowESS = 0.4)"},{"path":"/reference/check_mcmc.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Diagnostics of the MCMC — check_mcmc","text":"stan_fit stanfit object. n_draws Number MCMC draws. threshold_lowESS number [0,1] indicating minimum acceptable value relative ESS. See details.","code":""},{"path":"/reference/check_mcmc.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Diagnostics of the MCMC — check_mcmc","text":"warning message case detected problems.","code":""},{"path":"/reference/check_mcmc.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Diagnostics of the MCMC — check_mcmc","text":"Performs checks quality MCMC. See check_ESS() check_hmc_diagn() details.","code":""},{"path":"/reference/compute_sigma.html","id":null,"dir":"Reference","previous_headings":"","what":"Compute covariance matrix for some reference-based methods (JR, CIR) — compute_sigma","title":"Compute covariance matrix for some reference-based methods (JR, CIR) — compute_sigma","text":"Adapt covariance matrix reference-based methods. Used Copy Increments Reference (CIR) Jump Reference (JTR) methods, adapt covariance matrix different pre-deviation post deviation covariance structures. See Carpenter et al. (2013)","code":""},{"path":"/reference/compute_sigma.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Compute covariance matrix for some reference-based methods (JR, CIR) — compute_sigma","text":"","code":"compute_sigma(sigma_group, sigma_ref, index_mar)"},{"path":"/reference/compute_sigma.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Compute covariance matrix for some reference-based methods (JR, CIR) — compute_sigma","text":"sigma_group covariance matrix dimensions equal index_mar subjects original group sigma_ref covariance matrix dimensions equal index_mar subjects reference group index_mar logical vector indicating visits meet MAR assumption subject. .e. identifies observations non-MAR intercurrent event (ICE).","code":""},{"path":"/reference/compute_sigma.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Compute covariance matrix for some reference-based methods (JR, CIR) — compute_sigma","text":"Carpenter, James R., James H. Roger, Michael G. Kenward. \"Analysis longitudinal trials protocol deviation: framework relevant, accessible assumptions, inference via multiple imputation.\" Journal Biopharmaceutical statistics 23.6 (2013): 1352-1371.","code":""},{"path":"/reference/convert_to_imputation_list_df.html","id":null,"dir":"Reference","previous_headings":"","what":"Convert list of imputation_list_single() objects to an imputation_list_df() object\n(i.e. a list of imputation_df() objects's) — convert_to_imputation_list_df","title":"Convert list of imputation_list_single() objects to an imputation_list_df() object\n(i.e. a list of imputation_df() objects's) — convert_to_imputation_list_df","text":"Convert list imputation_list_single() objects imputation_list_df() object (.e. list imputation_df() objects's)","code":""},{"path":"/reference/convert_to_imputation_list_df.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Convert list of imputation_list_single() objects to an imputation_list_df() object\n(i.e. a list of imputation_df() objects's) — convert_to_imputation_list_df","text":"","code":"convert_to_imputation_list_df(imputes, sample_ids)"},{"path":"/reference/convert_to_imputation_list_df.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Convert list of imputation_list_single() objects to an imputation_list_df() object\n(i.e. a list of imputation_df() objects's) — convert_to_imputation_list_df","text":"imputes list imputation_list_single() objects sample_ids list 1 element per required imputation_df. element must contain vector \"ID\"'s correspond imputation_single() ID's required dataset. total number ID's must equal total number rows within imputes$imputations accommodate method_bmlmi() impute_data_individual() function returns list imputation_list_single() objects 1 object per subject. imputation_list_single() stores subjects imputations matrix columns matrix correspond D method_bmlmi(). Note methods (.e. methods_*()) special case D = 1. number rows matrix varies subject equal number times patient selected imputation (non-conditional mean methods 1 per subject per imputed dataset). function best illustrated example: convert_to_imputation_df(imputes, sample_ids) result : Note different repetitions (.e. value set D) grouped together sequentially.","code":"imputes = list( imputation_list_single( id = \"Tom\", imputations = matrix( imputation_single_t_1_1, imputation_single_t_1_2, imputation_single_t_2_1, imputation_single_t_2_2, imputation_single_t_3_1, imputation_single_t_3_2 ) ), imputation_list_single( id = \"Tom\", imputations = matrix( imputation_single_h_1_1, imputation_single_h_1_2, ) ) ) sample_ids <- list( c(\"Tom\", \"Harry\", \"Tom\"), c(\"Tom\") ) imputation_list_df( imputation_df( imputation_single_t_1_1, imputation_single_h_1_1, imputation_single_t_2_1 ), imputation_df( imputation_single_t_1_2, imputation_single_h_1_2, imputation_single_t_2_2 ), imputation_df( imputation_single_t_3_1 ), imputation_df( imputation_single_t_3_2 ) )"},{"path":"/reference/d_lagscale.html","id":null,"dir":"Reference","previous_headings":"","what":"Calculate delta from a lagged scale coefficient — d_lagscale","title":"Calculate delta from a lagged scale coefficient — d_lagscale","text":"Calculates delta value based upon baseline delta value post ICE scaling coefficient.","code":""},{"path":"/reference/d_lagscale.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Calculate delta from a lagged scale coefficient — d_lagscale","text":"","code":"d_lagscale(delta, dlag, is_post_ice)"},{"path":"/reference/d_lagscale.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Calculate delta from a lagged scale coefficient — d_lagscale","text":"delta numeric vector. Determines baseline amount delta applied visit. dlag numeric vector. Determines scaling applied delta based upon visit ICE occurred . Must length delta. is_post_ice logical vector. Indicates whether visit \"post-ICE\" .","code":""},{"path":"/reference/d_lagscale.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Calculate delta from a lagged scale coefficient — d_lagscale","text":"See delta_template() full details calculation performed.","code":""},{"path":"/reference/delta_template.html","id":null,"dir":"Reference","previous_headings":"","what":"Create a delta data.frame template — delta_template","title":"Create a delta data.frame template — delta_template","text":"Creates data.frame format required analyse() use applying delta adjustment.","code":""},{"path":"/reference/delta_template.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create a delta data.frame template — delta_template","text":"","code":"delta_template(imputations, delta = NULL, dlag = NULL, missing_only = TRUE)"},{"path":"/reference/delta_template.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create a delta data.frame template — delta_template","text":"imputations imputation object created impute(). delta NULL numeric vector. Determines baseline amount delta applied visit. See details. numeric vector must length number unique visits original dataset. dlag NULL numeric vector. Determines scaling applied delta based upon visit ICE occurred . See details. numeric vector must length number unique visits original dataset. missing_only Logical, TRUE non-missing post-ICE data delta value 0 assigned. Note calculation (described details section) performed first overwritten 0's end (.e. delta values missing post-ICE visits stay regardless option).","code":""},{"path":"/reference/delta_template.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Create a delta data.frame template — delta_template","text":"apply delta adjustment analyse() function expects delta data.frame 3 variables: vars$subjid, vars$visit delta (vars object supplied original call draws() created set_vars() function). function return data.frame aforementioned variables one row per subject per visit. delta argument function NULL delta column returned data.frame 0 observations. delta argument NULL delta calculated separately subject accumulative sum delta multiplied scaling coefficient dlag based upon many visits subject's intercurrent event (ICE) visit question . best illustrated example: Let delta = c(5,6,7,8) dlag=c(1,2,3,4) (.e. assuming 4 visits) lets say subject ICE visit 2. calculation follows: say subject delta offset 0 applied visit-1, 6 visit-2, 20 visit-3 44 visit-4. comparison, lets say subject instead ICE visit 3, calculation follows: terms practical usage, lets say wanted delta 5 used post ICE visits regardless proximity ICE visit. can achieved setting delta = c(5,5,5,5) dlag = c(1,0,0,0). example lets say subject ICE visit-1, calculation follows: Another way using arguments set delta difference time visits dlag amount delta per unit time. example lets say visit weeks 1, 5, 6 & 9 want delta 3 applied week ICE. can achieved setting delta = c(0,4,1,3) (difference weeks visit) dlag = c(3, 3, 3, 3). example lets say subject ICE week-5 (.e. visit-2) calculation : .e. week-6 (1 week ICE) delta 3 week-9 (4 weeks ICE) delta 12. Please note function also returns several utility variables user can create custom logic defining delta set . additional variables include: is_mar - observation missing regarded MAR? variable set FALSE observations occurred non-MAR ICE, otherwise set TRUE. is_missing - outcome variable observation missing. is_post_ice - observation occur patient's ICE defined data_ice dataset supplied draws(). strategy - imputation strategy assigned subject. design implementation function largely based upon functionality implemented called \"five marcos\" James Roger. See Roger (2021).","code":"v1 v2 v3 v4 -------------- 5 6 7 8 # delta assigned to each visit 0 1 2 3 # lagged scaling starting from the first visit after the subjects ICE -------------- 0 6 14 24 # delta * lagged scaling -------------- 0 6 20 44 # accumulative sum of delta to be applied to each visit v1 v2 v3 v4 -------------- 5 6 7 8 # delta assigned to each visit 0 0 1 2 # lagged scaling starting from the first visit after the subjects ICE -------------- 0 0 7 16 # delta * lagged scaling -------------- 0 0 7 23 # accumulative sum of delta to be applied to each visit v1 v2 v3 v4 -------------- 5 5 5 5 # delta assigned to each visit 1 0 0 0 # lagged scaling starting from the first visit after the subjects ICE -------------- 5 0 0 0 # delta * lagged scaling -------------- 5 5 5 5 # accumulative sum of delta to be applied to each visit v1 v2 v3 v4 -------------- 0 4 1 3 # delta assigned to each visit 0 0 3 3 # lagged scaling starting from the first visit after the subjects ICE -------------- 0 0 3 9 # delta * lagged scaling -------------- 0 0 3 12 # accumulative sum of delta to be applied to each visit"},{"path":"/reference/delta_template.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Create a delta data.frame template — delta_template","text":"Roger, James. Reference-based mi via multivariate normal rm (“five macros” miwithd), 2021. URL https://www.lshtm.ac.uk/research/centres-projects-groups/missing-data#dia-missing-data.","code":""},{"path":[]},{"path":"/reference/delta_template.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Create a delta data.frame template — delta_template","text":"","code":"if (FALSE) { delta_template(imputeObj) delta_template(imputeObj, delta = c(5,6,7,8), dlag = c(1,2,3,4)) }"},{"path":"/reference/do_not_run.html","id":null,"dir":"Reference","previous_headings":"","what":"Do not run this function — do_not_run","title":"Do not run this function — do_not_run","text":"function exists suppress false positive R CMD Check unused libraries","code":""},{"path":"/reference/do_not_run.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Do not run this function — do_not_run","text":"","code":"do_not_run()"},{"path":"/reference/do_not_run.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Do not run this function — do_not_run","text":"rstantools RcppParallel required used installation time. case RcppParallel used src/Makevars file created fly installation rstantools. rstantools used configure file.","code":""},{"path":"/reference/draws.html","id":null,"dir":"Reference","previous_headings":"","what":"Fit the base imputation model and get parameter estimates — draws","title":"Fit the base imputation model and get parameter estimates — draws","text":"draws fits base imputation model observed outcome data according given multiple imputation methodology. According user's method specification, returns either draws posterior distribution model parameters required Bayesian multiple imputation frequentist parameter estimates original data bootstrapped leave-one-datasets required conditional mean imputation. purpose imputation model estimate model parameters absence intercurrent events (ICEs) handled using reference-based imputation methods. reason, observed outcome data ICEs, reference-based imputation methods specified, removed considered missing purpose estimating imputation model, purpose . imputation model mixed model repeated measures (MMRM) valid missing--random (MAR) assumption. can fit using maximum likelihood (ML) restricted ML (REML) estimation, Bayesian approach, approximate Bayesian approach according user's method specification. ML/REML approaches approximate Bayesian approach support several possible covariance structures, Bayesian approach based MCMC sampling supports unstructured covariance structure. case covariance matrix can assumed different across group.","code":""},{"path":"/reference/draws.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Fit the base imputation model and get parameter estimates — draws","text":"","code":"draws(data, data_ice = NULL, vars, method, ncores = 1, quiet = FALSE) # S3 method for approxbayes draws(data, data_ice = NULL, vars, method, ncores = 1, quiet = FALSE) # S3 method for condmean draws(data, data_ice = NULL, vars, method, ncores = 1, quiet = FALSE) # S3 method for bmlmi draws(data, data_ice = NULL, vars, method, ncores = 1, quiet = FALSE) # S3 method for bayes draws(data, data_ice = NULL, vars, method, ncores = 1, quiet = FALSE)"},{"path":"/reference/draws.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Fit the base imputation model and get parameter estimates — draws","text":"data data.frame containing data used model. See details. data_ice data.frame specifies information related ICEs imputation strategies. See details. vars vars object generated set_vars(). See details. method method object generated either method_bayes(), method_approxbayes(), method_condmean() method_bmlmi(). specifies multiple imputation methodology used. See details. ncores single numeric specifying number cores use creating draws object. Note parameter ignored method_bayes() (Default = 1). quiet Logical, TRUE suppress printing progress information printed console.","code":""},{"path":"/reference/draws.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Fit the base imputation model and get parameter estimates — draws","text":"draws object named list containing following: data: R6 longdata object containing relevant input data information. method: method object generated either method_bayes(), method_approxbayes() method_condmean(). samples: list containing estimated parameters interest. element samples named list containing following: ids: vector characters containing ids subjects included original dataset. beta: numeric vector estimated regression coefficients. sigma: list estimated covariance matrices (one level vars$group). theta: numeric vector transformed covariances. failed: Logical. TRUE model fit failed. ids_samp: vector characters containing ids subjects included given sample. fit: method_bayes() chosen, returns MCMC Stan fit object. Otherwise NULL. n_failures: absolute number failures model fit. Relevant method_condmean(type = \"bootstrap\"), method_approxbayes() method_bmlmi(). formula: fixed effects formula object used model specification.","code":""},{"path":"/reference/draws.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Fit the base imputation model and get parameter estimates — draws","text":"draws performs first step multiple imputation (MI) procedure: fitting base imputation model. goal estimate parameters interest needed imputation phase (.e. regression coefficients covariance matrices MMRM model). function distinguishes following methods: Bayesian MI based MCMC sampling: draws returns draws posterior distribution parameters using Bayesian approach based MCMC sampling. method can specified using method = method_bayes(). Approximate Bayesian MI based bootstrapping: draws returns draws posterior distribution parameters using approximate Bayesian approach, sampling posterior distribution simulated fitting MMRM model bootstrap samples original dataset. method can specified using method = method_approxbayes()]. Conditional mean imputation bootstrap re-sampling: draws returns MMRM parameter estimates original dataset n_samples bootstrap samples. method can specified using method = method_condmean() argument type = \"bootstrap\". Conditional mean imputation jackknife re-sampling: draws returns MMRM parameter estimates original dataset leave-one-subject-sample. method can specified using method = method_condmean() argument type = \"jackknife\". Bootstrapped Maximum Likelihood MI: draws returns MMRM parameter estimates given number bootstrap samples needed perform random imputations bootstrapped samples. method can specified using method = method_bmlmi(). Bayesian MI based MCMC sampling proposed Carpenter, Roger, Kenward (2013) first introduced reference-based imputation methods. Approximate Bayesian MI discussed Little Rubin (2002). Conditional mean imputation methods discussed Wolbers et al (2022). Bootstrapped Maximum Likelihood MI described Von Hippel & Bartlett (2021). argument data contains longitudinal data. must least following variables: subjid: factor vector containing subject ids. visit: factor vector containing visit outcome observed . group: factor vector containing group subject belongs . outcome: numeric vector containing outcome variable. might contain missing values. Additional baseline time-varying covariates must included data. data must one row per visit per subject. means incomplete outcome data must set NA instead related row missing. Missing values covariates allowed. data incomplete expand_locf() helper function can used insert missing rows using Last Observation Carried Forward (LOCF) imputation impute covariates values. Note LOCF generally principled imputation method used appropriate specific covariate. Please note special provisioning baseline outcome values. want baseline observations included model part response variable removed advance outcome variable data. time want include baseline outcome covariate model, included separate column data (covariate). Character covariates explicitly cast factors. use custom analysis function requires specific reference levels character covariates (example computation least square means computation) advised manually cast character covariates factor advance running draws(). argument data_ice contains information occurrence ICEs. data.frame 3 columns: Subject ID: character vector containing ids subjects experienced ICE. column must named specified vars$subjid. Visit: character vector containing first visit occurrence ICE (.e. first visit affected ICE). visits must equal one levels data[[vars$visit]]. multiple ICEs happen subject, first non-MAR visit used. column must named specified vars$visit. Strategy: character vector specifying imputation strategy address ICE subject. column must named specified vars$strategy. Possible imputation strategies : \"MAR\": Missing Random. \"CIR\": Copy Increments Reference. \"CR\": Copy Reference. \"JR\": Jump Reference. \"LMCF\": Last Mean Carried Forward. explanations imputation strategies, see Carpenter, Roger, Kenward (2013), Cro et al (2021), Wolbers et al (2022). Please note user-defined imputation strategies can also set. data_ice argument necessary stage since (explained Wolbers et al (2022)), model fitted removing observations incompatible imputation model, .e. observed data data_ice[[vars$visit]] addressed imputation strategy different MAR excluded model fit. However observations discarded data imputation phase (performed function (impute()). summarize, stage pre-ICE data post-ICE data ICEs MAR imputation specified used. data_ice argument omitted, subject record within data_ice, assumed relevant subject's data pre-ICE missing visits imputed MAR assumption observed data used fit base imputation model. Please note ICE visit updated via update_strategy argument impute(); means subjects record data_ice always missing data imputed MAR assumption even strategy updated. vars argument named list specifies names key variables within data data_ice. list created set_vars() contains following named elements: subjid: name column data data_ice contains subject ids variable. visit: name column data data_ice contains visit variable. group: name column data contains group variable. outcome: name column data contains outcome variable. covariates: vector characters contains covariates included model (including interactions specified \"covariateName1*covariateName2\"``). covariates provided default model specification outcome ~ 1 + visit + groupwill used. Please note thegroup*visit` interaction included model default. strata: covariates used stratification variables bootstrap sampling. default vars$group set stratification variable. Needed method_condmean(type = \"bootstrap\") method_approxbayes(). strategy: name column data_ice contains subject-specific imputation strategy.","code":""},{"path":"/reference/draws.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Fit the base imputation model and get parameter estimates — draws","text":"James R Carpenter, James H Roger, Michael G Kenward. Analysis longitudinal trials protocol deviation: framework relevant, accessible assumptions, inference via multiple imputation. Journal Biopharmaceutical Statistics, 23(6):1352–1371, 2013. Suzie Cro, Tim P Morris, Michael G Kenward, James R Carpenter. Sensitivity analysis clinical trials missing continuous outcome data using controlled multiple imputation: practical guide. Statistics Medicine, 39(21):2815–2842, 2020. Roderick J. . Little Donald B. Rubin. Statistical Analysis Missing Data, Second Edition. John Wiley & Sons, Hoboken, New Jersey, 2002. [Section 10.2.3] Marcel Wolbers, Alessandro Noci, Paul Delmar, Craig Gower-Page, Sean Yiu, Jonathan W. Bartlett. Standard reference-based conditional mean imputation. https://arxiv.org/abs/2109.11162, 2022. Von Hippel, Paul T Bartlett, Jonathan W. Maximum likelihood multiple imputation: Faster imputations consistent standard errors without posterior draws. 2021.","code":""},{"path":[]},{"path":"/reference/encap_get_mmrm_sample.html","id":null,"dir":"Reference","previous_headings":"","what":"Encapsulate get_mmrm_sample — encap_get_mmrm_sample","title":"Encapsulate get_mmrm_sample — encap_get_mmrm_sample","text":"Function creates new wrapper function around get_mmrm_sample() arguments get_mmrm_sample() enclosed within new function. makes running parallel single process calls function smoother. particular function takes care exporting arguments required parallel process cluster","code":""},{"path":"/reference/encap_get_mmrm_sample.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Encapsulate get_mmrm_sample — encap_get_mmrm_sample","text":"","code":"encap_get_mmrm_sample(cl, longdata, method)"},{"path":"/reference/encap_get_mmrm_sample.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Encapsulate get_mmrm_sample — encap_get_mmrm_sample","text":"cl Either cluster get_cluster() NULL longdata longdata object longDataConstructor$new() method method object","code":""},{"path":[]},{"path":"/reference/eval_mmrm.html","id":null,"dir":"Reference","previous_headings":"","what":"Evaluate a call to mmrm — eval_mmrm","title":"Evaluate a call to mmrm — eval_mmrm","text":"utility function attempts evaluate call mmrm managing warnings errors thrown. particular function attempts catch warnings errors instead surfacing simply add additional element failed value TRUE. allows multiple calls made without program exiting.","code":""},{"path":"/reference/eval_mmrm.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Evaluate a call to mmrm — eval_mmrm","text":"","code":"eval_mmrm(expr)"},{"path":"/reference/eval_mmrm.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Evaluate a call to mmrm — eval_mmrm","text":"expr expression evaluated. call mmrm::mmrm().","code":""},{"path":"/reference/eval_mmrm.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Evaluate a call to mmrm — eval_mmrm","text":"function originally developed use glmmTMB needed hand-holding dropping false-positive warnings. important now kept around encase need catch false-positive warnings future.","code":""},{"path":[]},{"path":"/reference/eval_mmrm.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Evaluate a call to mmrm — eval_mmrm","text":"","code":"if (FALSE) { eval_mmrm({ mmrm::mmrm(formula, data) }) }"},{"path":"/reference/expand.html","id":null,"dir":"Reference","previous_headings":"","what":"Expand and fill in missing data.frame rows — expand","title":"Expand and fill in missing data.frame rows — expand","text":"functions essentially wrappers around base::expand.grid() ensure missing combinations data inserted data.frame imputation/fill methods updating covariate values newly created rows.","code":""},{"path":"/reference/expand.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Expand and fill in missing data.frame rows — expand","text":"","code":"expand(data, ...) fill_locf(data, vars, group = NULL, order = NULL) expand_locf(data, ..., vars, group, order)"},{"path":"/reference/expand.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Expand and fill in missing data.frame rows — expand","text":"data dataset expand fill . ... variables levels expanded (note duplicate entries levels result multiple rows level). vars character vector containing names variables need filled . group character vector containing names variables group performing LOCF imputation var. order character vector containing names additional variables sort data.frame performing LOCF.","code":""},{"path":"/reference/expand.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Expand and fill in missing data.frame rows — expand","text":"draws() function makes assumption subjects visits present data.frame covariate values non missing; expand(), fill_locf() expand_locf() utility functions support users ensuring data.frame's conform assumptions. expand() takes vectors expected levels data.frame expands combinations inserting missing rows data.frame. Note \"expanded\" variables cast factors. fill_locf() applies LOCF imputation named covariates fill NAs created insertion new rows expand() (though note distinction made existing NAs newly created NAs). Note data.frame sorted c(group, order) performing LOCF imputation; data.frame returned original sort order however. expand_locf() simple composition function fill_locf() expand() .e. fill_locf(expand(...)).","code":""},{"path":"/reference/expand.html","id":"missing-first-values","dir":"Reference","previous_headings":"","what":"Missing First Values","title":"Expand and fill in missing data.frame rows — expand","text":"fill_locf() function performs last observation carried forward imputation. natural consequence unable impute missing observations observation first value given subject / grouping. values deliberately imputed risks silent errors case time varying covariates. One solution first use expand_locf() just visit variable time varying covariates merge baseline covariates afterwards .e.","code":"library(dplyr) dat_expanded <- expand( data = dat, subject = c(\"pt1\", \"pt2\", \"pt3\", \"pt4\"), visit = c(\"vis1\", \"vis2\", \"vis3\") ) dat_filled <- dat_expanded %>% left_join(baseline_covariates, by = \"subject\")"},{"path":"/reference/expand.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Expand and fill in missing data.frame rows — expand","text":"","code":"if (FALSE) { dat_expanded <- expand( data = dat, subject = c(\"pt1\", \"pt2\", \"pt3\", \"pt4\"), visit = c(\"vis1\", \"vis2\", \"vis3\") ) dat_filled <- fill_loc( data = dat_expanded, vars = c(\"Sex\", \"Age\"), group = \"subject\", order = \"visit\" ) ## Or dat_filled <- expand_locf( data = dat, subject = c(\"pt1\", \"pt2\", \"pt3\", \"pt4\"), visit = c(\"vis1\", \"vis2\", \"vis3\"), vars = c(\"Sex\", \"Age\"), group = \"subject\", order = \"visit\" ) }"},{"path":"/reference/extract_covariates.html","id":null,"dir":"Reference","previous_headings":"","what":"Extract Variables from string vector — extract_covariates","title":"Extract Variables from string vector — extract_covariates","text":"Takes string including potentially model terms like * : extracts individual variables","code":""},{"path":"/reference/extract_covariates.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Extract Variables from string vector — extract_covariates","text":"","code":"extract_covariates(x)"},{"path":"/reference/extract_covariates.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Extract Variables from string vector — extract_covariates","text":"x string variable names potentially including interaction terms","code":""},{"path":"/reference/extract_covariates.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Extract Variables from string vector — extract_covariates","text":".e. c(\"v1\", \"v2\", \"v2*v3\", \"v1:v2\") becomes c(\"v1\", \"v2\", \"v3\")","code":""},{"path":"/reference/extract_data_nmar_as_na.html","id":null,"dir":"Reference","previous_headings":"","what":"Set to NA outcome values that would be MNAR if they were missing\n(i.e. which occur after an ICE handled using a reference-based imputation strategy) — extract_data_nmar_as_na","title":"Set to NA outcome values that would be MNAR if they were missing\n(i.e. which occur after an ICE handled using a reference-based imputation strategy) — extract_data_nmar_as_na","text":"Set NA outcome values MNAR missing (.e. occur ICE handled using reference-based imputation strategy)","code":""},{"path":"/reference/extract_data_nmar_as_na.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Set to NA outcome values that would be MNAR if they were missing\n(i.e. which occur after an ICE handled using a reference-based imputation strategy) — extract_data_nmar_as_na","text":"","code":"extract_data_nmar_as_na(longdata)"},{"path":"/reference/extract_data_nmar_as_na.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Set to NA outcome values that would be MNAR if they were missing\n(i.e. which occur after an ICE handled using a reference-based imputation strategy) — extract_data_nmar_as_na","text":"longdata R6 longdata object containing relevant input data information.","code":""},{"path":"/reference/extract_data_nmar_as_na.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Set to NA outcome values that would be MNAR if they were missing\n(i.e. which occur after an ICE handled using a reference-based imputation strategy) — extract_data_nmar_as_na","text":"data.frame containing longdata$get_data(longdata$ids), MNAR outcome values set NA.","code":""},{"path":"/reference/extract_draws.html","id":null,"dir":"Reference","previous_headings":"","what":"Extract draws from a stanfit object — extract_draws","title":"Extract draws from a stanfit object — extract_draws","text":"Extract draws stanfit object convert lists. function rstan::extract() returns draws given parameter array. function calls rstan::extract() extract draws stanfit object convert arrays lists.","code":""},{"path":"/reference/extract_draws.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Extract draws from a stanfit object — extract_draws","text":"","code":"extract_draws(stan_fit)"},{"path":"/reference/extract_draws.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Extract draws from a stanfit object — extract_draws","text":"stan_fit stanfit object.","code":""},{"path":"/reference/extract_draws.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Extract draws from a stanfit object — extract_draws","text":"named list length 2 containing: beta: list length equal number draws containing draws posterior distribution regression coefficients. sigma: list length equal number draws containing draws posterior distribution covariance matrices. element list list length equal 1 same_cov = TRUE equal number groups same_cov = FALSE.","code":""},{"path":"/reference/extract_imputed_df.html","id":null,"dir":"Reference","previous_headings":"","what":"Extract imputed dataset — extract_imputed_df","title":"Extract imputed dataset — extract_imputed_df","text":"Takes imputation object generated imputation_df() uses extract completed dataset longdata object created longDataConstructor(). Also applies delta transformation data.frame provided delta argument. See analyse() details structure data.frame. Subject IDs returned data.frame scrambled .e. original values.","code":""},{"path":"/reference/extract_imputed_df.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Extract imputed dataset — extract_imputed_df","text":"","code":"extract_imputed_df(imputation, ld, delta = NULL, idmap = FALSE)"},{"path":"/reference/extract_imputed_df.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Extract imputed dataset — extract_imputed_df","text":"imputation imputation object generated imputation_df(). ld longdata object generated longDataConstructor(). delta Either NULL data.frame. used offset outcome values imputed dataset. idmap Logical. TRUE attribute called \"idmap\" attached return object contains list maps old subject ids new subject ids.","code":""},{"path":"/reference/extract_imputed_df.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Extract imputed dataset — extract_imputed_df","text":"data.frame.","code":""},{"path":"/reference/extract_imputed_dfs.html","id":null,"dir":"Reference","previous_headings":"","what":"Extract imputed datasets — extract_imputed_dfs","title":"Extract imputed datasets — extract_imputed_dfs","text":"Extracts imputed datasets contained within imputations object generated impute().","code":""},{"path":"/reference/extract_imputed_dfs.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Extract imputed datasets — extract_imputed_dfs","text":"","code":"extract_imputed_dfs( imputations, index = seq_along(imputations$imputations), delta = NULL, idmap = FALSE )"},{"path":"/reference/extract_imputed_dfs.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Extract imputed datasets — extract_imputed_dfs","text":"imputations imputations object created impute(). index indexes imputed datasets return. default, datasets within imputations object returned. delta data.frame containing delta transformation applied imputed dataset. See analyse() details format specification data.frame. idmap Logical. subject IDs imputed data.frame's replaced new IDs ensure unique. Setting argument TRUE attaches attribute, called idmap, returned data.frame's provide map new subject IDs old subject IDs.","code":""},{"path":"/reference/extract_imputed_dfs.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Extract imputed datasets — extract_imputed_dfs","text":"list data.frames equal length index argument.","code":""},{"path":[]},{"path":"/reference/extract_imputed_dfs.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Extract imputed datasets — extract_imputed_dfs","text":"","code":"if (FALSE) { extract_imputed_dfs(imputeObj) extract_imputed_dfs(imputeObj, c(1:3)) }"},{"path":"/reference/extract_params.html","id":null,"dir":"Reference","previous_headings":"","what":"Extract parameters from a MMRM model — extract_params","title":"Extract parameters from a MMRM model — extract_params","text":"Extracts beta sigma coefficients MMRM model created mmrm::mmrm().","code":""},{"path":"/reference/extract_params.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Extract parameters from a MMRM model — extract_params","text":"","code":"extract_params(fit)"},{"path":"/reference/extract_params.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Extract parameters from a MMRM model — extract_params","text":"fit object created mmrm::mmrm()","code":""},{"path":"/reference/fit_mcmc.html","id":null,"dir":"Reference","previous_headings":"","what":"Fit the base imputation model using a Bayesian approach — fit_mcmc","title":"Fit the base imputation model using a Bayesian approach — fit_mcmc","text":"fit_mcmc() fits base imputation model using Bayesian approach. done MCMC method implemented stan run using function rstan::sampling(). function returns draws posterior distribution model parameters stanfit object. Additionally performs multiple diagnostics checks chain returns warnings case detected issues.","code":""},{"path":"/reference/fit_mcmc.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Fit the base imputation model using a Bayesian approach — fit_mcmc","text":"","code":"fit_mcmc(designmat, outcome, group, subjid, visit, method, quiet = FALSE)"},{"path":"/reference/fit_mcmc.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Fit the base imputation model using a Bayesian approach — fit_mcmc","text":"designmat design matrix fixed effects. outcome response variable. Must numeric. group Character vector containing group variable. subjid Character vector containing subjects IDs. visit Character vector containing visit variable. method method object generated method_bayes(). quiet Specify whether stan sampling log printed console.","code":""},{"path":"/reference/fit_mcmc.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Fit the base imputation model using a Bayesian approach — fit_mcmc","text":"named list composed following: samples: named list containing draws parameter. corresponds output extract_draws(). fit: stanfit object.","code":""},{"path":"/reference/fit_mcmc.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Fit the base imputation model using a Bayesian approach — fit_mcmc","text":"Bayesian model assumes multivariate normal likelihood function weakly-informative priors model parameters: particular, uniform priors assumed regression coefficients inverse-Wishart priors covariance matrices. chain initialized using REML parameter estimates MMRM starting values. function performs following steps: Fit MMRM using REML approach. Prepare input data MCMC fit described data{} block Stan file. See prepare_stan_data() details. Run MCMC according input arguments using starting values REML parameter estimates estimated point 1. Performs diagnostics checks MCMC. See check_mcmc() details. Extract draws model fit. chains perform method$n_samples draws keeping one every method$burn_between iterations. Additionally first method$burn_in iterations discarded. total number iterations method$burn_in + method$burn_between*method$n_samples. purpose method$burn_in ensure samples drawn stationary distribution Markov Chain. method$burn_between aims keep draws uncorrelated .","code":""},{"path":"/reference/fit_mmrm.html","id":null,"dir":"Reference","previous_headings":"","what":"Fit a MMRM model — fit_mmrm","title":"Fit a MMRM model — fit_mmrm","text":"Fits MMRM model allowing different covariance structures using mmrm::mmrm(). Returns list key model parameters beta, sigma additional element failed indicating whether fit failed converge. fit fail converge beta sigma present.","code":""},{"path":"/reference/fit_mmrm.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Fit a MMRM model — fit_mmrm","text":"","code":"fit_mmrm( designmat, outcome, subjid, visit, group, cov_struct = c(\"us\", \"toep\", \"cs\", \"ar1\"), REML = TRUE, same_cov = TRUE )"},{"path":"/reference/fit_mmrm.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Fit a MMRM model — fit_mmrm","text":"designmat data.frame matrix containing covariates use MMRM model. Dummy variables must already expanded , .e. via stats::model.matrix(). contain missing values outcome numeric vector. outcome value regressed MMRM model. subjid character / factor vector. subject identifier used link separate visits belong subject. visit character / factor vector. Indicates visit outcome value occurred . group character / factor vector. Indicates treatment group patient belongs . cov_struct character value. Specifies covariance structure use. Must one \"us\", \"toep\", \"cs\" \"ar1\" REML logical. Specifies whether restricted maximum likelihood used same_cov logical. Used specify shared individual covariance matrix used per group","code":""},{"path":"/reference/generate_data_single.html","id":null,"dir":"Reference","previous_headings":"","what":"Generate data for a single group — generate_data_single","title":"Generate data for a single group — generate_data_single","text":"Generate data single group","code":""},{"path":"/reference/generate_data_single.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Generate data for a single group — generate_data_single","text":"","code":"generate_data_single(pars_group, strategy_fun = NULL, distr_pars_ref = NULL)"},{"path":"/reference/generate_data_single.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Generate data for a single group — generate_data_single","text":"pars_group simul_pars object generated set_simul_pars(). specifies simulation parameters given group. strategy_fun Function implementing trajectories intercurrent event (ICE). Must one getStrategies(). See getStrategies() details. NULL post-ICE outcomes untouched. distr_pars_ref Optional. Named list containing simulation parameters reference arm. contains following elements: mu: Numeric vector indicating mean outcome trajectory assuming ICEs. include outcome baseline. sigma Covariance matrix outcome trajectory assuming ICEs. NULL, parameters inherited pars_group.","code":""},{"path":"/reference/generate_data_single.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Generate data for a single group — generate_data_single","text":"data.frame containing simulated data. includes following variables: id: Factor variable specifies id subject. visit: Factor variable specifies visit assessment. Visit 0 denotes baseline visit. group: Factor variable specifies treatment group subject belongs . outcome_bl: Numeric variable specifies baseline outcome. outcome_noICE: Numeric variable specifies longitudinal outcome assuming ICEs. ind_ice1: Binary variable takes value 1 corresponding visit affected ICE1 0 otherwise. dropout_ice1: Binary variable takes value 1 corresponding visit affected drop-following ICE1 0 otherwise. ind_ice2: Binary variable takes value 1 corresponding visit affected ICE2. outcome: Numeric variable specifies longitudinal outcome including ICE1, ICE2 intermittent missing values.","code":""},{"path":[]},{"path":"/reference/getStrategies.html","id":null,"dir":"Reference","previous_headings":"","what":"Get imputation strategies — getStrategies","title":"Get imputation strategies — getStrategies","text":"Returns list defining imputation strategies used create multivariate normal distribution parameters merging source group reference group per patient.","code":""},{"path":"/reference/getStrategies.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Get imputation strategies — getStrategies","text":"","code":"getStrategies(...)"},{"path":"/reference/getStrategies.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Get imputation strategies — getStrategies","text":"... User defined methods added return list. Input must function.","code":""},{"path":"/reference/getStrategies.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Get imputation strategies — getStrategies","text":"default Jump Reference (JR), Copy Reference (CR), Copy Increments Reference (CIR), Last Mean Carried Forward (LMCF) Missing Random (MAR) defined. user can define strategy functions (overwrite pre-defined ones) specifying named input function .e. NEW = function(...) .... exception MAR overwritten. user defined functions must take 3 inputs: pars_group, pars_ref index_mar. pars_group pars_ref lists elements mu sigma representing multivariate normal distribution parameters subject's current group reference group respectively. index_mar logical vector specifying visits subject met MAR assumption . function must return list elements mu sigma. See implementation strategy_JR() example.","code":""},{"path":"/reference/getStrategies.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Get imputation strategies — getStrategies","text":"","code":"if (FALSE) { getStrategies() getStrategies( NEW = function(pars_group, pars_ref, index_mar) code , JR = function(pars_group, pars_ref, index_mar) more_code ) }"},{"path":"/reference/get_ESS.html","id":null,"dir":"Reference","previous_headings":"","what":"Extract the Effective Sample Size (ESS) from a stanfit object — get_ESS","title":"Extract the Effective Sample Size (ESS) from a stanfit object — get_ESS","text":"Extract Effective Sample Size (ESS) stanfit object","code":""},{"path":"/reference/get_ESS.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Extract the Effective Sample Size (ESS) from a stanfit object — get_ESS","text":"","code":"get_ESS(stan_fit)"},{"path":"/reference/get_ESS.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Extract the Effective Sample Size (ESS) from a stanfit object — get_ESS","text":"stan_fit stanfit object.","code":""},{"path":"/reference/get_ESS.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Extract the Effective Sample Size (ESS) from a stanfit object — get_ESS","text":"named vector containing ESS parameter model.","code":""},{"path":"/reference/get_bootstrap_stack.html","id":null,"dir":"Reference","previous_headings":"","what":"Creates a stack object populated with bootstrapped samples — get_bootstrap_stack","title":"Creates a stack object populated with bootstrapped samples — get_bootstrap_stack","text":"Function creates Stack() object populated stack bootstrap samples based upon method$n_samples","code":""},{"path":"/reference/get_bootstrap_stack.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Creates a stack object populated with bootstrapped samples — get_bootstrap_stack","text":"","code":"get_bootstrap_stack(longdata, method, stack = Stack$new())"},{"path":"/reference/get_bootstrap_stack.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Creates a stack object populated with bootstrapped samples — get_bootstrap_stack","text":"longdata longDataConstructor() object method method object stack Stack() object (exposed unit testing purposes)","code":""},{"path":"/reference/get_cluster.html","id":null,"dir":"Reference","previous_headings":"","what":"Create cluster — get_cluster","title":"Create cluster — get_cluster","text":"Create cluster","code":""},{"path":"/reference/get_cluster.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create cluster — get_cluster","text":"","code":"get_cluster(ncores = 1)"},{"path":"/reference/get_cluster.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create cluster — get_cluster","text":"ncores Number parallel processes use ncores 1 function return NULL function spawns PSOCK cluster. Ensures rbmi assert_that loaded sub-processes","code":""},{"path":"/reference/get_conditional_parameters.html","id":null,"dir":"Reference","previous_headings":"","what":"Derive conditional multivariate normal parameters — get_conditional_parameters","title":"Derive conditional multivariate normal parameters — get_conditional_parameters","text":"Takes parameters multivariate normal distribution observed values calculate conditional distribution unobserved values.","code":""},{"path":"/reference/get_conditional_parameters.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Derive conditional multivariate normal parameters — get_conditional_parameters","text":"","code":"get_conditional_parameters(pars, values)"},{"path":"/reference/get_conditional_parameters.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Derive conditional multivariate normal parameters — get_conditional_parameters","text":"pars list elements mu sigma defining mean vector covariance matrix respectively. values vector observed values condition , must length pars$mu. Missing values must represented NA.","code":""},{"path":"/reference/get_conditional_parameters.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Derive conditional multivariate normal parameters — get_conditional_parameters","text":"list conditional distribution parameters: mu - conditional mean vector. sigma - conditional covariance matrix.","code":""},{"path":"/reference/get_delta_template.html","id":null,"dir":"Reference","previous_headings":"","what":"Get delta utility variables — get_delta_template","title":"Get delta utility variables — get_delta_template","text":"function creates default delta template (1 row per subject per visit) extracts utility information users need define logic defining delta. See delta_template() full details.","code":""},{"path":"/reference/get_delta_template.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Get delta utility variables — get_delta_template","text":"","code":"get_delta_template(imputations)"},{"path":"/reference/get_delta_template.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Get delta utility variables — get_delta_template","text":"imputations imputations object created impute().","code":""},{"path":"/reference/get_draws_mle.html","id":null,"dir":"Reference","previous_headings":"","what":"Fit the base imputation model on bootstrap samples — get_draws_mle","title":"Fit the base imputation model on bootstrap samples — get_draws_mle","text":"Fit base imputation model using ML/REML approach given number bootstrap samples specified method$n_samples. Returns parameter estimates model fit.","code":""},{"path":"/reference/get_draws_mle.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Fit the base imputation model on bootstrap samples — get_draws_mle","text":"","code":"get_draws_mle( longdata, method, sample_stack, n_target_samples, first_sample_orig, use_samp_ids, failure_limit = 0, ncores = 1, quiet = FALSE )"},{"path":"/reference/get_draws_mle.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Fit the base imputation model on bootstrap samples — get_draws_mle","text":"longdata R6 longdata object containing relevant input data information. method method object generated either method_approxbayes() method_condmean() argument type = \"bootstrap\". sample_stack stack object containing subject ids used mmrm iteration. n_target_samples Number samples needed created first_sample_orig Logical. TRUE function returns method$n_samples + 1 samples first sample contains parameter estimates original dataset method$n_samples samples contain parameter estimates bootstrap samples. FALSE function returns method$n_samples samples containing parameter estimates bootstrap samples. use_samp_ids Logical. TRUE, sampled subject ids returned. Otherwise subject ids original dataset returned. values used tell impute() subjects used derive imputed dataset. failure_limit Number failed samples allowed throwing error ncores Number processes parallelise job quiet Logical, TRUE suppress printing progress information printed console.","code":""},{"path":"/reference/get_draws_mle.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Fit the base imputation model on bootstrap samples — get_draws_mle","text":"draws object named list containing following: data: R6 longdata object containing relevant input data information. method: method object generated either method_bayes(), method_approxbayes() method_condmean(). samples: list containing estimated parameters interest. element samples named list containing following: ids: vector characters containing ids subjects included original dataset. beta: numeric vector estimated regression coefficients. sigma: list estimated covariance matrices (one level vars$group). theta: numeric vector transformed covariances. failed: Logical. TRUE model fit failed. ids_samp: vector characters containing ids subjects included given sample. fit: method_bayes() chosen, returns MCMC Stan fit object. Otherwise NULL. n_failures: absolute number failures model fit. Relevant method_condmean(type = \"bootstrap\"), method_approxbayes() method_bmlmi(). formula: fixed effects formula object used model specification.","code":""},{"path":"/reference/get_draws_mle.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Fit the base imputation model on bootstrap samples — get_draws_mle","text":"function takes Stack object contains multiple lists patient ids. function takes Stack pulls set ids constructs dataset just consisting patients (.e. potentially bootstrap jackknife sample). function fits MMRM model dataset create sample object. function repeats process n_target_samples reached. failure_limit samples fail converge function throws error. reaching desired number samples function generates returns draws object.","code":""},{"path":"/reference/get_ests_bmlmi.html","id":null,"dir":"Reference","previous_headings":"","what":"Von Hippel and Bartlett pooling of BMLMI method — get_ests_bmlmi","title":"Von Hippel and Bartlett pooling of BMLMI method — get_ests_bmlmi","text":"Compute pooled point estimates, standard error degrees freedom according Von Hippel Bartlett formula Bootstrapped Maximum Likelihood Multiple Imputation (BMLMI).","code":""},{"path":"/reference/get_ests_bmlmi.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Von Hippel and Bartlett pooling of BMLMI method — get_ests_bmlmi","text":"","code":"get_ests_bmlmi(ests, D)"},{"path":"/reference/get_ests_bmlmi.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Von Hippel and Bartlett pooling of BMLMI method — get_ests_bmlmi","text":"ests numeric vector containing estimates analysis imputed datasets. D numeric representing number imputations bootstrap sample BMLMI method.","code":""},{"path":"/reference/get_ests_bmlmi.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Von Hippel and Bartlett pooling of BMLMI method — get_ests_bmlmi","text":"list containing point estimate, standard error degrees freedom.","code":""},{"path":"/reference/get_ests_bmlmi.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Von Hippel and Bartlett pooling of BMLMI method — get_ests_bmlmi","text":"ests must provided following order: firsts D elements related analyses random imputation one bootstrap sample. second set D elements (.e. D+1 2*D) related second bootstrap sample .","code":""},{"path":"/reference/get_ests_bmlmi.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Von Hippel and Bartlett pooling of BMLMI method — get_ests_bmlmi","text":"Von Hippel, Paul T Bartlett, Jonathan W8. Maximum likelihood multiple imputation: Faster imputations consistent standard errors without posterior draws. 2021","code":""},{"path":"/reference/get_example_data.html","id":null,"dir":"Reference","previous_headings":"","what":"Simulate a realistic example dataset — get_example_data","title":"Simulate a realistic example dataset — get_example_data","text":"Simulate realistic example dataset using simulate_data() hard-coded values input arguments.","code":""},{"path":"/reference/get_example_data.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Simulate a realistic example dataset — get_example_data","text":"","code":"get_example_data()"},{"path":"/reference/get_example_data.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Simulate a realistic example dataset — get_example_data","text":"get_example_data() simulates 1:1 randomized trial active drug (intervention) versus placebo (control) 100 subjects per group 6 post-baseline assessments (bi-monthly visits 12 months). One intercurrent event corresponding treatment discontinuation also simulated. Specifically, data simulated following assumptions: mean outcome trajectory placebo group increases linearly 50 baseline (visit 0) 60 visit 6, .e. slope 10 points/year. mean outcome trajectory intervention group identical placebo group visit 2. visit 2 onward, slope decreases 50% 5 points/year. covariance structure baseline follow-values groups implied random intercept slope model standard deviation 5 intercept slope, correlation 0.25. addition, independent residual error standard deviation 2.5 added assessment. probability study drug discontinuation visit calculated according logistic model depends observed outcome visit. Specifically, visit-wise discontinuation probability 2% 3% control intervention group, respectively, specified case observed outcome equal 50 (mean value baseline). odds discontinuation simulated increase +10% +1 point increase observed outcome. Study drug discontinuation simulated effect mean trajectory placebo group. intervention group, subjects discontinue follow slope mean trajectory placebo group time point onward. compatible copy increments reference (CIR) assumption. Study drop-study drug discontinuation visit occurs probability 50% leading missing outcome data time point onward.","code":""},{"path":[]},{"path":"/reference/get_jackknife_stack.html","id":null,"dir":"Reference","previous_headings":"","what":"Creates a stack object populated with jackknife samples — get_jackknife_stack","title":"Creates a stack object populated with jackknife samples — get_jackknife_stack","text":"Function creates Stack() object populated stack jackknife samples based upon","code":""},{"path":"/reference/get_jackknife_stack.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Creates a stack object populated with jackknife samples — get_jackknife_stack","text":"","code":"get_jackknife_stack(longdata, method, stack = Stack$new())"},{"path":"/reference/get_jackknife_stack.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Creates a stack object populated with jackknife samples — get_jackknife_stack","text":"longdata longDataConstructor() object method method object stack Stack() object (exposed unit testing purposes)","code":""},{"path":"/reference/get_mmrm_sample.html","id":null,"dir":"Reference","previous_headings":"","what":"Fit MMRM and returns parameter estimates — get_mmrm_sample","title":"Fit MMRM and returns parameter estimates — get_mmrm_sample","text":"get_mmrm_sample fits base imputation model using ML/REML approach. Returns parameter estimates fit.","code":""},{"path":"/reference/get_mmrm_sample.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Fit MMRM and returns parameter estimates — get_mmrm_sample","text":"","code":"get_mmrm_sample(ids, longdata, method)"},{"path":"/reference/get_mmrm_sample.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Fit MMRM and returns parameter estimates — get_mmrm_sample","text":"ids vector characters containing ids subjects. longdata R6 longdata object containing relevant input data information. method method object generated either method_approxbayes() method_condmean().","code":""},{"path":"/reference/get_mmrm_sample.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Fit MMRM and returns parameter estimates — get_mmrm_sample","text":"named list class sample_single. contains following: ids vector characters containing ids subjects included original dataset. beta numeric vector estimated regression coefficients. sigma list estimated covariance matrices (one level vars$group). theta numeric vector transformed covariances. failed logical. TRUE model fit failed. ids_samp vector characters containing ids subjects included given sample.","code":""},{"path":"/reference/get_pattern_groups.html","id":null,"dir":"Reference","previous_headings":"","what":"Determine patients missingness group — get_pattern_groups","title":"Determine patients missingness group — get_pattern_groups","text":"Takes design matrix multiple rows per subject returns dataset 1 row per subject new column pgroup indicating group patient belongs (based upon missingness pattern treatment group)","code":""},{"path":"/reference/get_pattern_groups.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Determine patients missingness group — get_pattern_groups","text":"","code":"get_pattern_groups(ddat)"},{"path":"/reference/get_pattern_groups.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Determine patients missingness group — get_pattern_groups","text":"ddat data.frame columns subjid, visit, group, is_avail","code":""},{"path":"/reference/get_pattern_groups.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Determine patients missingness group — get_pattern_groups","text":"column is_avail must character numeric 0 1","code":""},{"path":"/reference/get_pattern_groups_unique.html","id":null,"dir":"Reference","previous_headings":"","what":"Get Pattern Summary — get_pattern_groups_unique","title":"Get Pattern Summary — get_pattern_groups_unique","text":"Takes dataset pattern information creates summary dataset just 1 row per pattern","code":""},{"path":"/reference/get_pattern_groups_unique.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Get Pattern Summary — get_pattern_groups_unique","text":"","code":"get_pattern_groups_unique(patterns)"},{"path":"/reference/get_pattern_groups_unique.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Get Pattern Summary — get_pattern_groups_unique","text":"patterns data.frame columns pgroup, pattern group","code":""},{"path":"/reference/get_pattern_groups_unique.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Get Pattern Summary — get_pattern_groups_unique","text":"column pgroup must numeric vector indicating pattern group patient belongs column pattern must character string 0's 1's. must identical rows within pgroup column group must character / numeric vector indicating covariance group observation belongs . must identical within pgroup","code":""},{"path":"/reference/get_pool_components.html","id":null,"dir":"Reference","previous_headings":"","what":"Expected Pool Components — get_pool_components","title":"Expected Pool Components — get_pool_components","text":"Returns elements expected contained analyse object depending analysis method specified.","code":""},{"path":"/reference/get_pool_components.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Expected Pool Components — get_pool_components","text":"","code":"get_pool_components(x)"},{"path":"/reference/get_pool_components.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Expected Pool Components — get_pool_components","text":"x Character name analysis method, must one either \"rubin\", \"jackknife\", \"bootstrap\" \"bmlmi\".","code":""},{"path":"/reference/get_visit_distribution_parameters.html","id":null,"dir":"Reference","previous_headings":"","what":"Derive visit distribution parameters — get_visit_distribution_parameters","title":"Derive visit distribution parameters — get_visit_distribution_parameters","text":"Takes patient level data beta coefficients expands get patient specific estimate visit distribution parameters mu sigma. Returns values specific format expected downstream functions imputation process (namely list(list(mu = ..., sigma = ...), list(mu = ..., sigma = ...))).","code":""},{"path":"/reference/get_visit_distribution_parameters.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Derive visit distribution parameters — get_visit_distribution_parameters","text":"","code":"get_visit_distribution_parameters(dat, beta, sigma)"},{"path":"/reference/get_visit_distribution_parameters.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Derive visit distribution parameters — get_visit_distribution_parameters","text":"dat Patient level dataset, must 1 row per visit. Column order must order beta. number columns must match length beta beta List model beta coefficients. 1 element sample e.g. 3 samples models 4 beta coefficients argument form list( c(1,2,3,4) , c(5,6,7,8), c(9,10,11,12)). elements beta must length must length order dat. sigma List sigma. Must number entries beta.","code":""},{"path":"/reference/has_class.html","id":null,"dir":"Reference","previous_headings":"","what":"Does object have a class ? — has_class","title":"Does object have a class ? — has_class","text":"Utility function see object particular class. Useful know many classes object may .","code":""},{"path":"/reference/has_class.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Does object have a class ? — has_class","text":"","code":"has_class(x, cls)"},{"path":"/reference/has_class.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Does object have a class ? — has_class","text":"x object want check class . cls class want know .","code":""},{"path":"/reference/has_class.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Does object have a class ? — has_class","text":"TRUE object class. FALSE object class.","code":""},{"path":"/reference/ife.html","id":null,"dir":"Reference","previous_headings":"","what":"if else — ife","title":"if else — ife","text":"wrapper around () else() prevent unexpected interactions ifelse() factor variables","code":""},{"path":"/reference/ife.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"if else — ife","text":"","code":"ife(x, a, b)"},{"path":"/reference/ife.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"if else — ife","text":"x True / False value return True b value return False","code":""},{"path":"/reference/ife.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"if else — ife","text":"default ifelse() convert factor variables numeric values often undesirable. connivance function avoids problem","code":""},{"path":"/reference/imputation_df.html","id":null,"dir":"Reference","previous_headings":"","what":"Create a valid imputation_df object — imputation_df","title":"Create a valid imputation_df object — imputation_df","text":"Create valid imputation_df object","code":""},{"path":"/reference/imputation_df.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create a valid imputation_df object — imputation_df","text":"","code":"imputation_df(...)"},{"path":"/reference/imputation_df.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create a valid imputation_df object — imputation_df","text":"... list imputation_single.","code":""},{"path":"/reference/imputation_list_df.html","id":null,"dir":"Reference","previous_headings":"","what":"List of imputations_df — imputation_list_df","title":"List of imputations_df — imputation_list_df","text":"container multiple imputation_df's","code":""},{"path":"/reference/imputation_list_df.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"List of imputations_df — imputation_list_df","text":"","code":"imputation_list_df(...)"},{"path":"/reference/imputation_list_df.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"List of imputations_df — imputation_list_df","text":"... objects class imputation_df","code":""},{"path":"/reference/imputation_list_single.html","id":null,"dir":"Reference","previous_headings":"","what":"A collection of imputation_singles() grouped by a single subjid ID — imputation_list_single","title":"A collection of imputation_singles() grouped by a single subjid ID — imputation_list_single","text":"collection imputation_singles() grouped single subjid ID","code":""},{"path":"/reference/imputation_list_single.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"A collection of imputation_singles() grouped by a single subjid ID — imputation_list_single","text":"","code":"imputation_list_single(imputations, D = 1)"},{"path":"/reference/imputation_list_single.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"A collection of imputation_singles() grouped by a single subjid ID — imputation_list_single","text":"imputations list imputation_single() objects ordered repetitions grouped sequentially D number repetitions performed determines many columns imputation matrix constructor function create imputation_list_single object contains matrix imputation_single() objects grouped single id. matrix split D columns (.e. non-bmlmi methods always 1) id attribute determined extracting id attribute contributing imputation_single() objects. error throw multiple id detected","code":""},{"path":"/reference/imputation_single.html","id":null,"dir":"Reference","previous_headings":"","what":"Create a valid imputation_single object — imputation_single","title":"Create a valid imputation_single object — imputation_single","text":"Create valid imputation_single object","code":""},{"path":"/reference/imputation_single.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create a valid imputation_single object — imputation_single","text":"","code":"imputation_single(id, values)"},{"path":"/reference/imputation_single.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create a valid imputation_single object — imputation_single","text":"id character string specifying subject id. values numeric vector indicating imputed values.","code":""},{"path":"/reference/impute.html","id":null,"dir":"Reference","previous_headings":"","what":"Create imputed datasets — impute","title":"Create imputed datasets — impute","text":"impute() creates imputed datasets based upon data options specified call draws(). One imputed dataset created per \"sample\" created draws().","code":""},{"path":"/reference/impute.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create imputed datasets — impute","text":"","code":"impute( draws, references = NULL, update_strategy = NULL, strategies = getStrategies() ) # S3 method for random impute( draws, references = NULL, update_strategy = NULL, strategies = getStrategies() ) # S3 method for condmean impute( draws, references = NULL, update_strategy = NULL, strategies = getStrategies() )"},{"path":"/reference/impute.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create imputed datasets — impute","text":"draws draws object created draws(). references named vector. Identifies references used reference-based imputation methods. form c(\"Group1\" = \"Reference1\", \"Group2\" = \"Reference2\"). NULL (default), references assumed form c(\"Group1\" = \"Group1\", \"Group2\" = \"Group2\"). argument NULL imputation strategy (defined data_ice[[vars$strategy]] call draws) MAR set. update_strategy optional data.frame. Updates imputation method originally set via data_ice option draws(). See details section information. strategies named list functions. Defines imputation functions used. names list mirror values specified strategy column data_ice. Default = getStrategies(). See getStrategies() details.","code":""},{"path":"/reference/impute.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Create imputed datasets — impute","text":"impute() uses imputation model parameter estimates, generated draws(), first calculate marginal (multivariate normal) distribution subject's longitudinal outcome variable depending covariate values. subjects intercurrent events (ICEs) handled using non-MAR methods, marginal distribution updated depending time first visit affected ICE, chosen imputation strategy chosen reference group described Carpenter, Roger, Kenward (2013) . subject's imputation distribution used imputing missing values defined marginal distribution conditional observed outcome values. One dataset generated per set parameter estimates provided draws(). exact manner missing values imputed conditional imputation distribution depends method object provided draws(), particular: Bayes & Approximate Bayes: imputed dataset contains 1 row per subject & visit original dataset missing values imputed taking single random sample conditional imputation distribution. Conditional Mean: imputed dataset contains 1 row per subject & visit bootstrapped jackknife dataset used generate corresponding parameter estimates draws(). Missing values imputed using mean conditional imputation distribution. Please note first imputed dataset refers conditional mean imputation original dataset whereas subsequent imputed datasets refer conditional mean imputations bootstrap jackknife samples, respectively, original data. Bootstrapped Maximum Likelihood MI (BMLMI): performs D random imputations bootstrapped dataset used generate corresponding parameter estimates draws(). total number B*D imputed datasets provided, B number bootstrapped datasets. Missing values imputed taking random sample conditional imputation distribution. update_strategy argument can used update imputation strategy originally set via data_ice option draws(). avoids re-run draws() function changing imputation strategy certain circumstances (detailed ). data.frame provided update_strategy argument must contain two columns, one subject ID another imputation strategy, whose names defined vars argument specified call draws(). Please note argument allows update imputation strategy arguments time first visit affected ICE. key limitation functionality one can switch MAR non-MAR strategy (vice versa) subjects without observed post-ICE data. reason change affect whether post-ICE data included base imputation model (explained help draws()). example, subject ICE \"Visit 2\" observed/known values \"Visit 3\" function throw error one tries switch strategy MAR non-MAR strategy. contrast, switching non-MAR MAR strategy, whilst valid, raise warning usable data utilised imputation model.","code":""},{"path":"/reference/impute.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Create imputed datasets — impute","text":"James R Carpenter, James H Roger, Michael G Kenward. Analysis longitudinal trials protocol deviation: framework relevant, accessible assumptions, inference via multiple imputation. Journal Biopharmaceutical Statistics, 23(6):1352–1371, 2013. [Section 4.2 4.3]","code":""},{"path":"/reference/impute.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Create imputed datasets — impute","text":"","code":"if (FALSE) { impute( draws = drawobj, references = c(\"Trt\" = \"Placebo\", \"Placebo\" = \"Placebo\") ) new_strategy <- data.frame( subjid = c(\"Pt1\", \"Pt2\"), strategy = c(\"MAR\", \"JR\") ) impute( draws = drawobj, references = c(\"Trt\" = \"Placebo\", \"Placebo\" = \"Placebo\"), update_strategy = new_strategy ) }"},{"path":"/reference/impute_data_individual.html","id":null,"dir":"Reference","previous_headings":"","what":"Impute data for a single subject — impute_data_individual","title":"Impute data for a single subject — impute_data_individual","text":"function performs imputation single subject time implementing process detailed impute().","code":""},{"path":"/reference/impute_data_individual.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Impute data for a single subject — impute_data_individual","text":"","code":"impute_data_individual( id, index, beta, sigma, data, references, strategies, condmean, n_imputations = 1 )"},{"path":"/reference/impute_data_individual.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Impute data for a single subject — impute_data_individual","text":"id Character string identifying subject. index sample indexes subject belongs e.g c(1,1,1,2,2,4). beta list beta coefficients sample, .e. beta[[1]] set beta coefficients first sample. sigma list sigma coefficients sample split group .e. sigma[[1]][[\"\"]] give sigma coefficients group first sample. data longdata object created longDataConstructor() references named vector. Identifies references used generating imputed values. form c(\"Group\" = \"Reference\", \"Group\" = \"Reference\"). strategies named list functions. Defines imputation functions used. names list mirror values specified method column data_ice. Default = getStrategies(). See getStrategies() details. condmean Logical. TRUE impute using conditional mean values, FALSE impute taking random draw multivariate normal distribution. n_imputations condmean = FALSE numeric representing number random imputations performed sample. Default 1 (one random imputation per sample).","code":""},{"path":"/reference/impute_data_individual.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Impute data for a single subject — impute_data_individual","text":"Note function performs required imputations subject time. .e. subject included samples 1,3,5,9 imputations (using sample-dependent imputation model parameters) performed one step order avoid look subjects's covariates expanding design matrix multiple times (computationally expensive). function also supports subject belonging sample multiple times, .e. 1,1,2,3,5,5, typically occur bootstrapped datasets.","code":""},{"path":"/reference/impute_internal.html","id":null,"dir":"Reference","previous_headings":"","what":"Create imputed datasets — impute_internal","title":"Create imputed datasets — impute_internal","text":"work horse function implements functionality impute. See user level function impute() details.","code":""},{"path":"/reference/impute_internal.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create imputed datasets — impute_internal","text":"","code":"impute_internal( draws, references = NULL, update_strategy, strategies, condmean )"},{"path":"/reference/impute_internal.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create imputed datasets — impute_internal","text":"draws draws object created draws(). references named vector. Identifies references used reference-based imputation methods. form c(\"Group1\" = \"Reference1\", \"Group2\" = \"Reference2\"). NULL (default), references assumed form c(\"Group1\" = \"Group1\", \"Group2\" = \"Group2\"). argument NULL imputation strategy (defined data_ice[[vars$strategy]] call draws) MAR set. update_strategy optional data.frame. Updates imputation method originally set via data_ice option draws(). See details section information. strategies named list functions. Defines imputation functions used. names list mirror values specified strategy column data_ice. Default = getStrategies(). See getStrategies() details. condmean logical. TRUE impute using conditional mean values, values impute taking random draw multivariate normal distribution.","code":""},{"path":"/reference/impute_outcome.html","id":null,"dir":"Reference","previous_headings":"","what":"Sample outcome value — impute_outcome","title":"Sample outcome value — impute_outcome","text":"Draws random sample multivariate normal distribution.","code":""},{"path":"/reference/impute_outcome.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Sample outcome value — impute_outcome","text":"","code":"impute_outcome(conditional_parameters, n_imputations = 1, condmean = FALSE)"},{"path":"/reference/impute_outcome.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Sample outcome value — impute_outcome","text":"conditional_parameters list elements mu sigma contain mean vector covariance matrix sample . n_imputations numeric representing number random samples multivariate normal distribution performed. Default 1. condmean conditional mean imputation performed (opposed random sampling)","code":""},{"path":"/reference/invert.html","id":null,"dir":"Reference","previous_headings":"","what":"invert — invert","title":"invert — invert","text":"Utility function used replicated purrr::transpose. Turns list inside .","code":""},{"path":"/reference/invert.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"invert — invert","text":"","code":"invert(x)"},{"path":"/reference/invert.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"invert — invert","text":"x list","code":""},{"path":"/reference/invert_indexes.html","id":null,"dir":"Reference","previous_headings":"","what":"Invert and derive indexes — invert_indexes","title":"Invert and derive indexes — invert_indexes","text":"Takes list elements creates new list containing 1 entry per unique element value containing indexes original elements occurred .","code":""},{"path":"/reference/invert_indexes.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Invert and derive indexes — invert_indexes","text":"","code":"invert_indexes(x)"},{"path":"/reference/invert_indexes.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Invert and derive indexes — invert_indexes","text":"x list elements invert calculate index (see details).","code":""},{"path":"/reference/invert_indexes.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Invert and derive indexes — invert_indexes","text":"functions purpose best illustrated example: input: becomes:","code":"list( c(\"A\", \"B\", \"C\"), c(\"A\", \"A\", \"B\"))} list( \"A\" = c(1,2,2), \"B\" = c(1,2), \"C\" = 1 )"},{"path":"/reference/is_absent.html","id":null,"dir":"Reference","previous_headings":"","what":"Is value absent — is_absent","title":"Is value absent — is_absent","text":"Returns true value either NULL, NA \"\". case vector values must NULL/NA/\"\" x regarded absent.","code":""},{"path":"/reference/is_absent.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Is value absent — is_absent","text":"","code":"is_absent(x, na = TRUE, blank = TRUE)"},{"path":"/reference/is_absent.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Is value absent — is_absent","text":"x value check absent na NAs count absent blank blanks .e. \"\" count absent","code":""},{"path":"/reference/is_char_fact.html","id":null,"dir":"Reference","previous_headings":"","what":"Is character or factor — is_char_fact","title":"Is character or factor — is_char_fact","text":"returns true x character factor vector","code":""},{"path":"/reference/is_char_fact.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Is character or factor — is_char_fact","text":"","code":"is_char_fact(x)"},{"path":"/reference/is_char_fact.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Is character or factor — is_char_fact","text":"x character factor vector","code":""},{"path":"/reference/is_char_one.html","id":null,"dir":"Reference","previous_headings":"","what":"Is single character — is_char_one","title":"Is single character — is_char_one","text":"returns true x length 1 character vector","code":""},{"path":"/reference/is_char_one.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Is single character — is_char_one","text":"","code":"is_char_one(x)"},{"path":"/reference/is_char_one.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Is single character — is_char_one","text":"x character vector","code":""},{"path":"/reference/is_in_rbmi_development.html","id":null,"dir":"Reference","previous_headings":"","what":"Is package in development mode? — is_in_rbmi_development","title":"Is package in development mode? — is_in_rbmi_development","text":"Returns TRUE package developed .e. local copy source code actively editing Returns FALSE otherwise","code":""},{"path":"/reference/is_in_rbmi_development.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Is package in development mode? — is_in_rbmi_development","text":"","code":"is_in_rbmi_development()"},{"path":"/reference/is_in_rbmi_development.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Is package in development mode? — is_in_rbmi_development","text":"Main use function parallel processing indicate whether sub-processes need load current development version code whether load main installed package system","code":""},{"path":"/reference/is_num_char_fact.html","id":null,"dir":"Reference","previous_headings":"","what":"Is character, factor or numeric — is_num_char_fact","title":"Is character, factor or numeric — is_num_char_fact","text":"returns true x character, numeric factor vector","code":""},{"path":"/reference/is_num_char_fact.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Is character, factor or numeric — is_num_char_fact","text":"","code":"is_num_char_fact(x)"},{"path":"/reference/is_num_char_fact.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Is character, factor or numeric — is_num_char_fact","text":"x character, numeric factor vector","code":""},{"path":"/reference/locf.html","id":null,"dir":"Reference","previous_headings":"","what":"Last Observation Carried Forward — locf","title":"Last Observation Carried Forward — locf","text":"Returns vector applied last observation carried forward imputation.","code":""},{"path":"/reference/locf.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Last Observation Carried Forward — locf","text":"","code":"locf(x)"},{"path":"/reference/locf.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Last Observation Carried Forward — locf","text":"x vector.","code":""},{"path":"/reference/locf.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Last Observation Carried Forward — locf","text":"","code":"if (FALSE) { locf(c(NA, 1, 2, 3, NA, 4)) # Returns c(NA, 1, 2, 3, 3, 4) }"},{"path":"/reference/longDataConstructor.html","id":null,"dir":"Reference","previous_headings":"","what":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"longdata object allows efficient storage recall longitudinal datasets use bootstrap sampling. object works de-constructing data lists based upon subject id thus enabling efficient lookup.","code":""},{"path":"/reference/longDataConstructor.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"object also handles multiple operations specific rbmi defining whether outcome value MAR / Missing well tracking imputation strategy assigned subject. recognised objects functionality fairly overloaded hoped can split area specific objects / functions future. additions functionality object avoided possible.","code":""},{"path":"/reference/longDataConstructor.html","id":"public-fields","dir":"Reference","previous_headings":"","what":"Public fields","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"data original dataset passed constructor (sorted id visit) vars vars object (list key variables) passed constructor visits character vector containing distinct visit levels ids character vector containing unique ids subject self$data formula formula expressing design matrix data constructed strata numeric vector indicating strata corresponding value self$ids belongs . stratification variable defined default 1 subjects (.e. group). field used part self$sample_ids() function enable stratified bootstrap sampling ice_visit_index list indexed subject storing index number first visit affected ICE. ICE set equal number visits plus 1. values list indexed subject storing numeric vector original (unimputed) outcome values group list indexed subject storing single character indicating imputation group subject belongs defined self$data[id, self$ivars$group] used determine reference group used imputing subjects data. is_mar list indexed subject storing logical values indicating subjects outcome values MAR . list defaulted TRUE subjects & outcomes modified calls self$set_strategies(). Note indicate values missing, variable True outcome values either occurred ICE visit post ICE visit imputation strategy MAR strategies list indexed subject storing single character value indicating imputation strategy assigned subject. list defaulted \"MAR\" subjects modified calls either self$set_strategies() self$update_strategies() strategy_lock list indexed subject storing single logical value indicating whether patients imputation strategy locked . strategy locked means change MAR non-MAR. Strategies can changed non-MAR MAR though trigger warning. Strategies locked patient assigned MAR strategy non-missing ICE date. list populated call self$set_strategies(). indexes list indexed subject storing numeric vector indexes specify rows original dataset belong subject .e. recover full data subject \"pt3\" can use self$data[self$indexes[[\"pt3\"]],]. may seem redundant filtering data directly however enables efficient bootstrap sampling data .e. list populated object initialisation. is_missing list indexed subject storing logical vector indicating whether corresponding outcome subject missing. list populated object initialisation. is_post_ice list indexed subject storing logical vector indicating whether corresponding outcome subject post date ICE. ICE data provided defaults False observations. list populated call self$set_strategies().","code":"indexes <- unlist(self$indexes[c(\"pt3\", \"pt3\")]) self$data[indexes,]"},{"path":[]},{"path":"/reference/longDataConstructor.html","id":"public-methods","dir":"Reference","previous_headings":"","what":"Public methods","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"longDataConstructor$get_data() longDataConstructor$add_subject() longDataConstructor$validate_ids() longDataConstructor$sample_ids() longDataConstructor$extract_by_id() longDataConstructor$update_strategies() longDataConstructor$set_strategies() longDataConstructor$check_has_data_at_each_visit() longDataConstructor$set_strata() longDataConstructor$new() longDataConstructor$clone()","code":""},{"path":"/reference/longDataConstructor.html","id":"method-get-data-","dir":"Reference","previous_headings":"","what":"Method get_data()","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"Returns data.frame based upon required subject IDs. Replaces missing values new ones provided.","code":""},{"path":"/reference/longDataConstructor.html","id":"usage","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"","code":"longDataConstructor$get_data( obj = NULL, nmar.rm = FALSE, na.rm = FALSE, idmap = FALSE )"},{"path":"/reference/longDataConstructor.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"obj Either NULL, character vector subjects IDs imputation list object. See details. nmar.rm Logical value. TRUE remove observations regarded MAR (determined self$is_mar). na.rm Logical value. TRUE remove outcome values missing (determined self$is_missing). idmap Logical value. TRUE add attribute idmap contains mapping new subject ids old subject ids. See details.","code":""},{"path":"/reference/longDataConstructor.html","id":"details-1","dir":"Reference","previous_headings":"","what":"Details","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"obj NULL full original dataset returned. obj character vector new dataset consisting just subjects returned; character vector contains duplicate entries subject returned multiple times. obj imputation_df object (created imputation_df()) subject ids specified object returned missing values filled specified imputation list object. .e. return data.frame consisting observations pt1 twice observations pt3 . first set observations pt1 missing values filled c(1,2,3) second set filled c(4,5,6). length values must equal sum(self$is_missing[[id]]). obj NULL subject IDs scrambled order ensure unique .e. pt2 requested twice process guarantees set observations unique subject ID number. idmap attribute (requested) can used map new ids back old ids.","code":"obj <- imputation_df( imputation_single( id = \"pt1\", values = c(1,2,3)), imputation_single( id = \"pt1\", values = c(4,5,6)), imputation_single( id = \"pt3\", values = c(7,8)) ) longdata$get_data(obj)"},{"path":"/reference/longDataConstructor.html","id":"returns","dir":"Reference","previous_headings":"","what":"Returns","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"data.frame.","code":""},{"path":"/reference/longDataConstructor.html","id":"method-add-subject-","dir":"Reference","previous_headings":"","what":"Method add_subject()","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"function decomposes patient data self$data populates corresponding lists .e. self$is_missing, self$values, self$group, etc. function called upon objects initialization.","code":""},{"path":"/reference/longDataConstructor.html","id":"usage-1","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"","code":"longDataConstructor$add_subject(id)"},{"path":"/reference/longDataConstructor.html","id":"arguments-1","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"id Character subject id exists within self$data.","code":""},{"path":"/reference/longDataConstructor.html","id":"method-validate-ids-","dir":"Reference","previous_headings":"","what":"Method validate_ids()","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"Throws error element ids within source data self$data.","code":""},{"path":"/reference/longDataConstructor.html","id":"usage-2","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"","code":"longDataConstructor$validate_ids(ids)"},{"path":"/reference/longDataConstructor.html","id":"arguments-2","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"ids character vector ids.","code":""},{"path":"/reference/longDataConstructor.html","id":"returns-1","dir":"Reference","previous_headings":"","what":"Returns","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"TRUE","code":""},{"path":"/reference/longDataConstructor.html","id":"method-sample-ids-","dir":"Reference","previous_headings":"","what":"Method sample_ids()","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"Performs random stratified sampling patient ids (replacement) patient equal weight picked within strata (.e dependent many non-missing visits ).","code":""},{"path":"/reference/longDataConstructor.html","id":"usage-3","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"","code":"longDataConstructor$sample_ids()"},{"path":"/reference/longDataConstructor.html","id":"returns-2","dir":"Reference","previous_headings":"","what":"Returns","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"Character vector ids.","code":""},{"path":"/reference/longDataConstructor.html","id":"method-extract-by-id-","dir":"Reference","previous_headings":"","what":"Method extract_by_id()","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"Returns list key information given subject. convenience wrapper save manually grab element.","code":""},{"path":"/reference/longDataConstructor.html","id":"usage-4","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"","code":"longDataConstructor$extract_by_id(id)"},{"path":"/reference/longDataConstructor.html","id":"arguments-3","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"id Character subject id exists within self$data.","code":""},{"path":"/reference/longDataConstructor.html","id":"method-update-strategies-","dir":"Reference","previous_headings":"","what":"Method update_strategies()","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"Convenience function run self$set_strategies(dat_ice, update=TRUE) kept legacy reasons.","code":""},{"path":"/reference/longDataConstructor.html","id":"usage-5","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"","code":"longDataConstructor$update_strategies(dat_ice)"},{"path":"/reference/longDataConstructor.html","id":"arguments-4","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"dat_ice data.frame containing ICE information see impute() format dataframe.","code":""},{"path":"/reference/longDataConstructor.html","id":"method-set-strategies-","dir":"Reference","previous_headings":"","what":"Method set_strategies()","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"Updates self$strategies, self$is_mar, self$is_post_ice variables based upon provided ICE information.","code":""},{"path":"/reference/longDataConstructor.html","id":"usage-6","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"","code":"longDataConstructor$set_strategies(dat_ice = NULL, update = FALSE)"},{"path":"/reference/longDataConstructor.html","id":"arguments-5","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"dat_ice data.frame containing ICE information. See details. update Logical, indicates ICE data used update. See details.","code":""},{"path":"/reference/longDataConstructor.html","id":"details-2","dir":"Reference","previous_headings":"","what":"Details","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"See draws() specification dat_ice update=FALSE. See impute() format dat_ice update=TRUE. update=TRUE function ensures MAR strategies changed non-MAR presence post-ICE observations.","code":""},{"path":"/reference/longDataConstructor.html","id":"method-check-has-data-at-each-visit-","dir":"Reference","previous_headings":"","what":"Method check_has_data_at_each_visit()","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"Ensures visits least 1 observed \"MAR\" observation. Throws error criteria met. ensure initial MMRM can resolved.","code":""},{"path":"/reference/longDataConstructor.html","id":"usage-7","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"","code":"longDataConstructor$check_has_data_at_each_visit()"},{"path":"/reference/longDataConstructor.html","id":"method-set-strata-","dir":"Reference","previous_headings":"","what":"Method set_strata()","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"Populates self$strata variable. user specified stratification variables first visit used determine value variables. stratification variables specified everyone defined strata 1.","code":""},{"path":"/reference/longDataConstructor.html","id":"usage-8","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"","code":"longDataConstructor$set_strata()"},{"path":"/reference/longDataConstructor.html","id":"method-new-","dir":"Reference","previous_headings":"","what":"Method new()","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"Constructor function.","code":""},{"path":"/reference/longDataConstructor.html","id":"usage-9","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"","code":"longDataConstructor$new(data, vars)"},{"path":"/reference/longDataConstructor.html","id":"arguments-6","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"data longitudinal dataset. vars ivars object created set_vars().","code":""},{"path":"/reference/longDataConstructor.html","id":"method-clone-","dir":"Reference","previous_headings":"","what":"Method clone()","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"objects class cloneable method.","code":""},{"path":"/reference/longDataConstructor.html","id":"usage-10","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"","code":"longDataConstructor$clone(deep = FALSE)"},{"path":"/reference/longDataConstructor.html","id":"arguments-7","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for Storing / Accessing & Sampling Longitudinal Data — longDataConstructor","text":"deep Whether make deep clone.","code":""},{"path":"/reference/ls_design.html","id":null,"dir":"Reference","previous_headings":"","what":"Calculate design vector for the lsmeans — ls_design","title":"Calculate design vector for the lsmeans — ls_design","text":"Calculates design vector required generate lsmean standard error. ls_design_equal calculates applying equal weight per covariate combination whilst ls_design_proportional applies weighting proportional frequency covariate combination occurred actual dataset.","code":""},{"path":"/reference/ls_design.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Calculate design vector for the lsmeans — ls_design","text":"","code":"ls_design_equal(data, frm, covars, fix) ls_design_proportional(data, frm, covars, fix)"},{"path":"/reference/ls_design.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Calculate design vector for the lsmeans — ls_design","text":"data data.frame frm Formula used fit original model covars character vector variables names exist data extracted (ls_design_equal ) fix named list variables fixed values","code":""},{"path":"/reference/lsmeans.html","id":null,"dir":"Reference","previous_headings":"","what":"Least Square Means — lsmeans","title":"Least Square Means — lsmeans","text":"Estimates least square means linear model. done generating prediction model using hypothetical observation constructed averaging data. See details information.","code":""},{"path":"/reference/lsmeans.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Least Square Means — lsmeans","text":"","code":"lsmeans(model, ..., .weights = c(\"proportional\", \"equal\"))"},{"path":"/reference/lsmeans.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Least Square Means — lsmeans","text":"model model created lm. ... Fixes specific variables specific values .e. trt = 1 age = 50. name argument must name variable within dataset. .weights Character, specifies whether use \"proportional\" \"equal\" weighting categorical covariate combination calculating lsmeans.","code":""},{"path":"/reference/lsmeans.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Least Square Means — lsmeans","text":"lsmeans obtained calculating hypothetical patients predicting expected values. hypothetical patients constructed expanding possible combinations categorical covariate setting numerical covariates equal mean. final lsmean value calculated averaging hypothetical patients. .weights equals \"proportional\" values weighted frequency occur full dataset. .weights equals \"equal\" hypothetical patient given equal weight regardless actually occurs dataset. Use ... argument fix specific variables specific values. See references identical implementations done SAS R via emmeans package. function attempts re-implement emmeans derivation standard linear models without include dependencies.","code":""},{"path":"/reference/lsmeans.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Least Square Means — lsmeans","text":"https://CRAN.R-project.org/package=emmeans https://documentation.sas.com/doc/en/pgmsascdc/9.4_3.3/statug/statug_glm_details41.htm","code":""},{"path":"/reference/lsmeans.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Least Square Means — lsmeans","text":"","code":"if (FALSE) { mod <- lm(Sepal.Length ~ Species + Petal.Length, data = iris) lsmeans(mod) lsmeans(mod, Species = \"virginica\") lsmeans(mod, Species = \"versicolor\") lsmeans(mod, Species = \"versicolor\", Petal.Length = 1) }"},{"path":"/reference/method.html","id":null,"dir":"Reference","previous_headings":"","what":"Set the multiple imputation methodology — method","title":"Set the multiple imputation methodology — method","text":"functions determine methods rbmi use creating imputation models, generating imputed values pooling results.","code":""},{"path":"/reference/method.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Set the multiple imputation methodology — method","text":"","code":"method_bayes( burn_in = 200, burn_between = 50, same_cov = TRUE, n_samples = 20, seed = sample.int(.Machine$integer.max, 1) ) method_approxbayes( covariance = c(\"us\", \"toep\", \"cs\", \"ar1\"), threshold = 0.01, same_cov = TRUE, REML = TRUE, n_samples = 20 ) method_condmean( covariance = c(\"us\", \"toep\", \"cs\", \"ar1\"), threshold = 0.01, same_cov = TRUE, REML = TRUE, n_samples = NULL, type = c(\"bootstrap\", \"jackknife\") ) method_bmlmi( covariance = c(\"us\", \"toep\", \"cs\", \"ar1\"), threshold = 0.01, same_cov = TRUE, REML = TRUE, B = 20, D = 2 )"},{"path":"/reference/method.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Set the multiple imputation methodology — method","text":"burn_in numeric specifies many observations discarded prior extracting actual samples. Note sampler initialized maximum likelihood estimates weakly informative prior used thus theory value need high. burn_between numeric specifies \"thinning\" rate .e. many observations discarded sample. used prevent issues associated autocorrelation samples. same_cov logical, TRUE imputation model fitted using single shared covariance matrix observations. FALSE separate covariance matrix fit group determined group argument set_vars(). n_samples numeric determines many imputed datasets generated. case method_condmean(type = \"jackknife\") argument must set NULL. See details. seed numeric specifies seed used call Stan. argument passed onto seed argument rstan::sampling(). Note required method_bayes(), methods can achieve reproducible results setting seed via set.seed(). See details. covariance character string specifies structure covariance matrix used imputation model. Must one \"us\" (default), \"toep\", \"cs\" \"ar1\". See details. threshold numeric 0 1, specifies proportion bootstrap datasets can fail produce valid samples error thrown. See details. REML logical indicating whether use REML estimation rather maximum likelihood. type character string specifies resampling method used perform inference conditional mean imputation approach (set via method_condmean()) used. Must one \"bootstrap\" \"jackknife\". B numeric determines number bootstrap samples method_bmlmi. D numeric determines number random imputations bootstrap sample. Needed method_bmlmi().","code":""},{"path":"/reference/method.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Set the multiple imputation methodology — method","text":"case method_condmean(type = \"bootstrap\") n_samples + 1 imputation models datasets generated first sample based original dataset whilst n_samples samples bootstrapped datasets. Likewise, method_condmean(type = \"jackknife\") length(unique(data$subjid)) + 1 imputation models datasets generated. cases represented n + 1 displayed print message. user able specify different covariance structures using covariance argument. Currently supported structures include: Unstructured (\"us\") Toeplitz (\"toep\") Compound Symmetry (\"cs\") Autoregression-1 (\"ar1\") Note present Bayesian methods support unstructured. case method_condmean(type = \"bootstrap\"), method_approxbayes() method_bmlmi() repeated bootstrap samples original dataset taken MMRM fitted sample. Due randomness sampled datasets, well limitations optimisers used fit models, uncommon estimates particular dataset generated. instances rbmi designed throw bootstrapped dataset try another. However ensure errors due chance due underlying misspecification data /model tolerance limit set many samples can discarded. tolerance limit reached error thrown process aborted. tolerance limit defined ceiling(threshold * n_samples). Note jackknife method estimates need generated leave-one-datasets error thrown fail fit. Please note time writing (September 2021) Stan unable produce reproducible samples across different operating systems even seed used. care must taken using Stan across different machines. information limitation please consult Stan documentation https://mc-stan.org/docs/2_27/reference-manual/reproducibility-chapter.html","code":""},{"path":"/reference/parametric_ci.html","id":null,"dir":"Reference","previous_headings":"","what":"Calculate parametric confidence intervals — parametric_ci","title":"Calculate parametric confidence intervals — parametric_ci","text":"Calculates confidence intervals based upon parametric distribution.","code":""},{"path":"/reference/parametric_ci.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Calculate parametric confidence intervals — parametric_ci","text":"","code":"parametric_ci(point, se, alpha, alternative, qfun, pfun, ...)"},{"path":"/reference/parametric_ci.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Calculate parametric confidence intervals — parametric_ci","text":"point point estimate. se standard error point estimate. using non-\"normal\" distribution set 1. alpha type 1 error rate, value 0 1. alternative character string specifying alternative hypothesis, must one \"two.sided\" (default), \"greater\" \"less\". qfun quantile function assumed distribution .e. qnorm. pfun CDF function assumed distribution .e. pnorm. ... additional arguments passed qfun pfun .e. df = 102.","code":""},{"path":"/reference/pool.html","id":null,"dir":"Reference","previous_headings":"","what":"Pool analysis results obtained from the imputed datasets — pool","title":"Pool analysis results obtained from the imputed datasets — pool","text":"Pool analysis results obtained imputed datasets","code":""},{"path":"/reference/pool.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Pool analysis results obtained from the imputed datasets — pool","text":"","code":"pool( results, conf.level = 0.95, alternative = c(\"two.sided\", \"less\", \"greater\"), type = c(\"percentile\", \"normal\") ) # S3 method for pool as.data.frame(x, ...) # S3 method for pool print(x, ...)"},{"path":"/reference/pool.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Pool analysis results obtained from the imputed datasets — pool","text":"results analysis object created analyse(). conf.level confidence level returned confidence interval. Must single number 0 1. Default 0.95. alternative character string specifying alternative hypothesis, must one \"two.sided\" (default), \"greater\" \"less\". type character string either \"percentile\" (default) \"normal\". Determines method used calculate bootstrap confidence intervals. See details. used method_condmean(type = \"bootstrap\") specified original call draws(). x pool object generated pool(). ... used.","code":""},{"path":"/reference/pool.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Pool analysis results obtained from the imputed datasets — pool","text":"calculation used generate point estimate, standard errors confidence interval depends upon method specified original call draws(); particular: method_approxbayes() & method_bayes() use Rubin's rules pool estimates variances across multiple imputed datasets, Barnard-Rubin rule pool degree's freedom; see Little & Rubin (2002). method_condmean(type = \"bootstrap\") uses percentile normal approximation; see Efron & Tibshirani (1994). Note percentile bootstrap, standard error calculated, .e. standard errors NA object / data.frame. method_condmean(type = \"jackknife\") uses standard jackknife variance formula; see Efron & Tibshirani (1994). method_bmlmi uses pooling procedure Bootstrapped Maximum Likelihood MI (BMLMI). See Von Hippel & Bartlett (2021).","code":""},{"path":"/reference/pool.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Pool analysis results obtained from the imputed datasets — pool","text":"Bradley Efron Robert J Tibshirani. introduction bootstrap. CRC press, 1994. [Section 11] Roderick J. . Little Donald B. Rubin. Statistical Analysis Missing Data, Second Edition. John Wiley & Sons, Hoboken, New Jersey, 2002. [Section 5.4] Von Hippel, Paul T Bartlett, Jonathan W. Maximum likelihood multiple imputation: Faster imputations consistent standard errors without posterior draws. 2021.","code":""},{"path":"/reference/pool_bootstrap_normal.html","id":null,"dir":"Reference","previous_headings":"","what":"Bootstrap Pooling via normal approximation — pool_bootstrap_normal","title":"Bootstrap Pooling via normal approximation — pool_bootstrap_normal","text":"Get point estimate, confidence interval p-value using normal approximation.","code":""},{"path":"/reference/pool_bootstrap_normal.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Bootstrap Pooling via normal approximation — pool_bootstrap_normal","text":"","code":"pool_bootstrap_normal(est, conf.level, alternative)"},{"path":"/reference/pool_bootstrap_normal.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Bootstrap Pooling via normal approximation — pool_bootstrap_normal","text":"est numeric vector point estimates bootstrap sample. conf.level confidence level returned confidence interval. Must single number 0 1. Default 0.95. alternative character string specifying alternative hypothesis, must one \"two.sided\" (default), \"greater\" \"less\".","code":""},{"path":"/reference/pool_bootstrap_normal.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Bootstrap Pooling via normal approximation — pool_bootstrap_normal","text":"point estimate taken first element est. remaining n-1 values est used generate confidence intervals.","code":""},{"path":"/reference/pool_bootstrap_percentile.html","id":null,"dir":"Reference","previous_headings":"","what":"Bootstrap Pooling via Percentiles — pool_bootstrap_percentile","title":"Bootstrap Pooling via Percentiles — pool_bootstrap_percentile","text":"Get point estimate, confidence interval p-value using percentiles. Note quantile \"type=6\" used, see stats::quantile() details.","code":""},{"path":"/reference/pool_bootstrap_percentile.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Bootstrap Pooling via Percentiles — pool_bootstrap_percentile","text":"","code":"pool_bootstrap_percentile(est, conf.level, alternative)"},{"path":"/reference/pool_bootstrap_percentile.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Bootstrap Pooling via Percentiles — pool_bootstrap_percentile","text":"est numeric vector point estimates bootstrap sample. conf.level confidence level returned confidence interval. Must single number 0 1. Default 0.95. alternative character string specifying alternative hypothesis, must one \"two.sided\" (default), \"greater\" \"less\".","code":""},{"path":"/reference/pool_bootstrap_percentile.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Bootstrap Pooling via Percentiles — pool_bootstrap_percentile","text":"point estimate taken first element est. remaining n-1 values est used generate confidence intervals.","code":""},{"path":"/reference/pool_internal.html","id":null,"dir":"Reference","previous_headings":"","what":"Internal Pool Methods — pool_internal","title":"Internal Pool Methods — pool_internal","text":"Dispatches pool methods based upon results object class. See pool() details.","code":""},{"path":"/reference/pool_internal.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Internal Pool Methods — pool_internal","text":"","code":"pool_internal(results, conf.level, alternative, type, D) # S3 method for jackknife pool_internal(results, conf.level, alternative, type, D) # S3 method for bootstrap pool_internal( results, conf.level, alternative, type = c(\"percentile\", \"normal\"), D ) # S3 method for bmlmi pool_internal(results, conf.level, alternative, type, D) # S3 method for rubin pool_internal(results, conf.level, alternative, type, D)"},{"path":"/reference/pool_internal.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Internal Pool Methods — pool_internal","text":"results list results .e. x$results element analyse object created analyse()). conf.level confidence level returned confidence interval. Must single number 0 1. Default 0.95. alternative character string specifying alternative hypothesis, must one \"two.sided\" (default), \"greater\" \"less\". type character string either \"percentile\" (default) \"normal\". Determines method used calculate bootstrap confidence intervals. See details. used method_condmean(type = \"bootstrap\") specified original call draws(). D numeric representing number imputations bootstrap sample BMLMI method.","code":""},{"path":"/reference/prepare_stan_data.html","id":null,"dir":"Reference","previous_headings":"","what":"Prepare input data to run the Stan model — prepare_stan_data","title":"Prepare input data to run the Stan model — prepare_stan_data","text":"Prepare input data run Stan model. Creates / calculates required inputs required data{} block MMRM Stan program.","code":""},{"path":"/reference/prepare_stan_data.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Prepare input data to run the Stan model — prepare_stan_data","text":"","code":"prepare_stan_data(ddat, subjid, visit, outcome, group)"},{"path":"/reference/prepare_stan_data.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Prepare input data to run the Stan model — prepare_stan_data","text":"ddat design matrix subjid Character vector containing subjects IDs. visit Vector containing visits. outcome Numeric vector containing outcome variable. group Vector containing group variable.","code":""},{"path":"/reference/prepare_stan_data.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Prepare input data to run the Stan model — prepare_stan_data","text":"stan_data object. named list per data{} block related Stan file. particular returns: N - number rows design matrix P - number columns design matrix G - number distinct covariance matrix groups (.e. length(unique(group))) n_visit - number unique outcome visits n_pat - total number pattern groups (defined missingness patterns & covariance group) pat_G - Index Sigma pattern group use pat_n_pt - number patients within pattern group pat_n_visit - number non-missing visits pattern group pat_sigma_index - rows/cols Sigma subset pattern group (padded 0's) y - outcome variable Q - design matrix (QR decomposition) R - R matrix QR decomposition design matrix","code":""},{"path":"/reference/prepare_stan_data.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Prepare input data to run the Stan model — prepare_stan_data","text":"group argument determines covariance matrix group subject belongs . want subjects use shared covariance matrix set group \"1\" everyone.","code":""},{"path":"/reference/print.analysis.html","id":null,"dir":"Reference","previous_headings":"","what":"Print analysis object — print.analysis","title":"Print analysis object — print.analysis","text":"Print analysis object","code":""},{"path":"/reference/print.analysis.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Print analysis object — print.analysis","text":"","code":"# S3 method for analysis print(x, ...)"},{"path":"/reference/print.analysis.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Print analysis object — print.analysis","text":"x analysis object generated analyse(). ... used.","code":""},{"path":"/reference/print.draws.html","id":null,"dir":"Reference","previous_headings":"","what":"Print draws object — print.draws","title":"Print draws object — print.draws","text":"Print draws object","code":""},{"path":"/reference/print.draws.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Print draws object — print.draws","text":"","code":"# S3 method for draws print(x, ...)"},{"path":"/reference/print.draws.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Print draws object — print.draws","text":"x draws object generated draws(). ... used.","code":""},{"path":"/reference/print.imputation.html","id":null,"dir":"Reference","previous_headings":"","what":"Print imputation object — print.imputation","title":"Print imputation object — print.imputation","text":"Print imputation object","code":""},{"path":"/reference/print.imputation.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Print imputation object — print.imputation","text":"","code":"# S3 method for imputation print(x, ...)"},{"path":"/reference/print.imputation.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Print imputation object — print.imputation","text":"x imputation object generated impute(). ... used.","code":""},{"path":"/reference/progressLogger.html","id":null,"dir":"Reference","previous_headings":"","what":"R6 Class for printing current sampling progress — progressLogger","title":"R6 Class for printing current sampling progress — progressLogger","text":"Object initalised total number iterations expected occur. User can update object add method indicate many iterations just occurred. Every time step * 100 % iterations occurred message printed console. Use quiet argument prevent object printing anything ","code":""},{"path":"/reference/progressLogger.html","id":"public-fields","dir":"Reference","previous_headings":"","what":"Public fields","title":"R6 Class for printing current sampling progress — progressLogger","text":"step real, percentage iterations allow printing progress console step_current integer, total number iterations completed since progress last printed console n integer, current number completed iterations n_max integer, total number expected iterations completed acts denominator calculating progress percentages quiet logical holds whether print anything","code":""},{"path":[]},{"path":"/reference/progressLogger.html","id":"public-methods","dir":"Reference","previous_headings":"","what":"Public methods","title":"R6 Class for printing current sampling progress — progressLogger","text":"progressLogger$new() progressLogger$add() progressLogger$print_progress() progressLogger$clone()","code":""},{"path":"/reference/progressLogger.html","id":"method-new-","dir":"Reference","previous_headings":"","what":"Method new()","title":"R6 Class for printing current sampling progress — progressLogger","text":"Create progressLogger object","code":""},{"path":"/reference/progressLogger.html","id":"usage","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for printing current sampling progress — progressLogger","text":"","code":"progressLogger$new(n_max, quiet = FALSE, step = 0.1)"},{"path":"/reference/progressLogger.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for printing current sampling progress — progressLogger","text":"n_max integer, sets field n_max quiet logical, sets field quiet step real, sets field step","code":""},{"path":"/reference/progressLogger.html","id":"method-add-","dir":"Reference","previous_headings":"","what":"Method add()","title":"R6 Class for printing current sampling progress — progressLogger","text":"Records n iterations completed add number current step count (step_current) print progress message log step limit (step) reached. function nothing quiet set TRUE","code":""},{"path":"/reference/progressLogger.html","id":"usage-1","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for printing current sampling progress — progressLogger","text":"","code":"progressLogger$add(n)"},{"path":"/reference/progressLogger.html","id":"arguments-1","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for printing current sampling progress — progressLogger","text":"n number successfully complete iterations since add() last called","code":""},{"path":"/reference/progressLogger.html","id":"method-print-progress-","dir":"Reference","previous_headings":"","what":"Method print_progress()","title":"R6 Class for printing current sampling progress — progressLogger","text":"method print current state progress","code":""},{"path":"/reference/progressLogger.html","id":"usage-2","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for printing current sampling progress — progressLogger","text":"","code":"progressLogger$print_progress()"},{"path":"/reference/progressLogger.html","id":"method-clone-","dir":"Reference","previous_headings":"","what":"Method clone()","title":"R6 Class for printing current sampling progress — progressLogger","text":"objects class cloneable method.","code":""},{"path":"/reference/progressLogger.html","id":"usage-3","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for printing current sampling progress — progressLogger","text":"","code":"progressLogger$clone(deep = FALSE)"},{"path":"/reference/progressLogger.html","id":"arguments-2","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for printing current sampling progress — progressLogger","text":"deep Whether make deep clone.","code":""},{"path":"/reference/pval_percentile.html","id":null,"dir":"Reference","previous_headings":"","what":"P-value of percentile bootstrap — pval_percentile","title":"P-value of percentile bootstrap — pval_percentile","text":"Determines (necessarily unique) quantile (type=6) \"est\" gives value 0 , derive p-value corresponding percentile bootstrap via inversion.","code":""},{"path":"/reference/pval_percentile.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"P-value of percentile bootstrap — pval_percentile","text":"","code":"pval_percentile(est)"},{"path":"/reference/pval_percentile.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"P-value of percentile bootstrap — pval_percentile","text":"est numeric vector point estimates bootstrap sample.","code":""},{"path":"/reference/pval_percentile.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"P-value of percentile bootstrap — pval_percentile","text":"named numeric vector length 2 containing p-value H_0: theta=0 vs H_A: theta>0 (\"pval_greater\") p-value H_0: theta=0 vs H_A: theta<0 (\"pval_less\").","code":""},{"path":"/reference/pval_percentile.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"P-value of percentile bootstrap — pval_percentile","text":"p-value H_0: theta=0 vs H_A: theta>0 value alpha q_alpha = 0. least one estimate equal zero returns largest alpha q_alpha = 0. bootstrap estimates > 0 returns 0; bootstrap estimates < 0 returns 1. Analogous reasoning applied p-value H_0: theta=0 vs H_A: theta<0.","code":""},{"path":"/reference/random_effects_expr.html","id":null,"dir":"Reference","previous_headings":"","what":"Construct random effects formula — random_effects_expr","title":"Construct random effects formula — random_effects_expr","text":"Constructs character representation random effects formula fitting MMRM subject visit format required mmrm::mmrm().","code":""},{"path":"/reference/random_effects_expr.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Construct random effects formula — random_effects_expr","text":"","code":"random_effects_expr( cov_struct = c(\"us\", \"toep\", \"cs\", \"ar1\"), cov_by_group = FALSE )"},{"path":"/reference/random_effects_expr.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Construct random effects formula — random_effects_expr","text":"cov_struct Character - covariance structure used, must one \"us\", \"toep\", \"cs\", \"ar1\" cov_by_group Boolean - Whenever use separate covariances per group level","code":""},{"path":"/reference/random_effects_expr.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Construct random effects formula — random_effects_expr","text":"example assuming user specified covariance structure \"us\" groups provided return cov_by_group set FALSE indicates separate covariance matrices required per group following returned:","code":"us(visit | subjid) us( visit | group / subjid )"},{"path":"/reference/rbmi-package.html","id":null,"dir":"Reference","previous_headings":"","what":"rbmi: Reference Based Multiple Imputation — rbmi-package","title":"rbmi: Reference Based Multiple Imputation — rbmi-package","text":"rbmi package used perform reference based multiple imputation. package provides implementations common, patient-specific imputation strategies whilst allowing user select various standard Bayesian frequentist approaches. package designed around 4 core functions: draws() - Fits multiple imputation models impute() - Imputes multiple datasets analyse() - Analyses multiple datasets pool() - Pools multiple results single statistic learn rbmi, please see quickstart vignette: vignette(topic= \"quickstart\", package = \"rbmi\")","code":""},{"path":[]},{"path":"/reference/rbmi-package.html","id":"author","dir":"Reference","previous_headings":"","what":"Author","title":"rbmi: Reference Based Multiple Imputation — rbmi-package","text":"Maintainer: Craig Gower-Page craig.gower-page@roche.com Authors: Alessandro Noci alessandro.noci@roche.com contributors: Marcel Wolbers marcel.wolbers@roche.com [contributor] Roche [copyright holder, funder]","code":""},{"path":"/reference/record.html","id":null,"dir":"Reference","previous_headings":"","what":"Capture all Output — record","title":"Capture all Output — record","text":"function silences warnings, errors & messages instead returns list containing results (error) + warning error messages character vectors.","code":""},{"path":"/reference/record.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Capture all Output — record","text":"","code":"record(expr)"},{"path":"/reference/record.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Capture all Output — record","text":"expr expression executed","code":""},{"path":"/reference/record.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Capture all Output — record","text":"list containing results - object returned expr list() error thrown warnings - NULL character vector warnings thrown errors - NULL string error thrown messages - NULL character vector messages produced","code":""},{"path":"/reference/record.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Capture all Output — record","text":"","code":"if (FALSE) { record({ x <- 1 y <- 2 warning(\"something went wrong\") message(\"O nearly done\") x + y }) }"},{"path":"/reference/recursive_reduce.html","id":null,"dir":"Reference","previous_headings":"","what":"recursive_reduce — recursive_reduce","title":"recursive_reduce — recursive_reduce","text":"Utility function used replicated purrr::reduce. Recursively applies function list elements 1 element remains","code":""},{"path":"/reference/recursive_reduce.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"recursive_reduce — recursive_reduce","text":"","code":"recursive_reduce(.l, .f)"},{"path":"/reference/recursive_reduce.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"recursive_reduce — recursive_reduce","text":".l list values apply function .f function apply element list turn .e. .l[[1]] <- .f( .l[[1]] , .l[[2]]) ; .l[[1]] <- .f( .l[[1]] , .l[[3]])","code":""},{"path":"/reference/remove_if_all_missing.html","id":null,"dir":"Reference","previous_headings":"","what":"Remove subjects from dataset if they have no observed values — remove_if_all_missing","title":"Remove subjects from dataset if they have no observed values — remove_if_all_missing","text":"function takes data.frame variables visit, outcome & subjid. removes rows given subjid non-missing values outcome.","code":""},{"path":"/reference/remove_if_all_missing.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Remove subjects from dataset if they have no observed values — remove_if_all_missing","text":"","code":"remove_if_all_missing(dat)"},{"path":"/reference/remove_if_all_missing.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Remove subjects from dataset if they have no observed values — remove_if_all_missing","text":"dat data.frame","code":""},{"path":"/reference/rubin_df.html","id":null,"dir":"Reference","previous_headings":"","what":"Barnard and Rubin degrees of freedom adjustment — rubin_df","title":"Barnard and Rubin degrees of freedom adjustment — rubin_df","text":"Compute degrees freedom according Barnard-Rubin formula.","code":""},{"path":"/reference/rubin_df.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Barnard and Rubin degrees of freedom adjustment — rubin_df","text":"","code":"rubin_df(v_com, var_b, var_t, M)"},{"path":"/reference/rubin_df.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Barnard and Rubin degrees of freedom adjustment — rubin_df","text":"v_com Positive number representing degrees freedom complete-data analysis. var_b -variance point estimate across multiply imputed datasets. var_t Total-variance point estimate according Rubin's rules. M Number imputations.","code":""},{"path":"/reference/rubin_df.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Barnard and Rubin degrees of freedom adjustment — rubin_df","text":"Degrees freedom according Barnard-Rubin formula. See Barnard-Rubin (1999).","code":""},{"path":"/reference/rubin_df.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Barnard and Rubin degrees of freedom adjustment — rubin_df","text":"computation takes account limit cases missing data (.e. -variance var_b zero) complete-data degrees freedom set Inf. Moreover, v_com given NA, function returns Inf.","code":""},{"path":"/reference/rubin_df.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Barnard and Rubin degrees of freedom adjustment — rubin_df","text":"Barnard, J. Rubin, D.B. (1999). Small sample degrees freedom multiple imputation. Biometrika, 86, 948-955.","code":""},{"path":"/reference/rubin_rules.html","id":null,"dir":"Reference","previous_headings":"","what":"Combine estimates using Rubin's rules — rubin_rules","title":"Combine estimates using Rubin's rules — rubin_rules","text":"Pool together results M complete-data analyses according Rubin's rules. See details.","code":""},{"path":"/reference/rubin_rules.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Combine estimates using Rubin's rules — rubin_rules","text":"","code":"rubin_rules(ests, ses, v_com)"},{"path":"/reference/rubin_rules.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Combine estimates using Rubin's rules — rubin_rules","text":"ests Numeric vector containing point estimates complete-data analyses. ses Numeric vector containing standard errors complete-data analyses. v_com Positive number representing degrees freedom complete-data analysis.","code":""},{"path":"/reference/rubin_rules.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Combine estimates using Rubin's rules — rubin_rules","text":"list containing: est_point: pooled point estimate according Little-Rubin (2002). var_t: total variance according Little-Rubin (2002). df: degrees freedom according Barnard-Rubin (1999).","code":""},{"path":"/reference/rubin_rules.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Combine estimates using Rubin's rules — rubin_rules","text":"rubin_rules applies Rubin's rules (Rubin, 1987) pooling together results multiple imputation procedure. pooled point estimate est_point average across point estimates complete-data analyses (given input argument ests). total variance var_t sum two terms representing within-variance -variance (see Little-Rubin (2002)). function also returns df, estimated pooled degrees freedom according Barnard-Rubin (1999) can used inference based t-distribution.","code":""},{"path":"/reference/rubin_rules.html","id":"references","dir":"Reference","previous_headings":"","what":"References","title":"Combine estimates using Rubin's rules — rubin_rules","text":"Barnard, J. Rubin, D.B. (1999). Small sample degrees freedom multiple imputation. Biometrika, 86, 948-955 Roderick J. . Little Donald B. Rubin. Statistical Analysis Missing Data, Second Edition. John Wiley & Sons, Hoboken, New Jersey, 2002. [Section 5.4]","code":""},{"path":[]},{"path":"/reference/sample_ids.html","id":null,"dir":"Reference","previous_headings":"","what":"Sample Patient Ids — sample_ids","title":"Sample Patient Ids — sample_ids","text":"Performs stratified bootstrap sample IDS ensuring return vector length input vector","code":""},{"path":"/reference/sample_ids.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Sample Patient Ids — sample_ids","text":"","code":"sample_ids(ids, strata = rep(1, length(ids)))"},{"path":"/reference/sample_ids.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Sample Patient Ids — sample_ids","text":"ids vector sample strata strata indicator, ids sampled within strata ensuring numbers strata maintained","code":""},{"path":"/reference/sample_ids.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Sample Patient Ids — sample_ids","text":"","code":"if (FALSE) { sample_ids( c(\"a\", \"b\", \"c\", \"d\"), strata = c(1,1,2,2)) }"},{"path":"/reference/sample_list.html","id":null,"dir":"Reference","previous_headings":"","what":"Create and validate a sample_list object — sample_list","title":"Create and validate a sample_list object — sample_list","text":"Given list sample_single objects generate sample_single(), creates sample_list objects validate .","code":""},{"path":"/reference/sample_list.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create and validate a sample_list object — sample_list","text":"","code":"sample_list(...)"},{"path":"/reference/sample_list.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create and validate a sample_list object — sample_list","text":"... list sample_single objects.","code":""},{"path":"/reference/sample_mvnorm.html","id":null,"dir":"Reference","previous_headings":"","what":"Sample random values from the multivariate normal distribution — sample_mvnorm","title":"Sample random values from the multivariate normal distribution — sample_mvnorm","text":"Sample random values multivariate normal distribution","code":""},{"path":"/reference/sample_mvnorm.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Sample random values from the multivariate normal distribution — sample_mvnorm","text":"","code":"sample_mvnorm(mu, sigma)"},{"path":"/reference/sample_mvnorm.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Sample random values from the multivariate normal distribution — sample_mvnorm","text":"mu mean vector sigma covariance matrix Samples multivariate normal variables multiplying univariate random normal variables cholesky decomposition covariance matrix. mu length 1 just uses rnorm instead.","code":""},{"path":"/reference/sample_single.html","id":null,"dir":"Reference","previous_headings":"","what":"Create object of sample_single class — sample_single","title":"Create object of sample_single class — sample_single","text":"Creates object class sample_single named list containing input parameters validate .","code":""},{"path":"/reference/sample_single.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create object of sample_single class — sample_single","text":"","code":"sample_single( ids, beta = NA, sigma = NA, theta = NA, failed = any(is.na(beta)), ids_samp = ids )"},{"path":"/reference/sample_single.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create object of sample_single class — sample_single","text":"ids Vector characters containing ids subjects included original dataset. beta Numeric vector estimated regression coefficients. sigma List estimated covariance matrices (one level vars$group). theta Numeric vector transformed covariances. failed Logical. TRUE model fit failed. ids_samp Vector characters containing ids subjects included given sample.","code":""},{"path":"/reference/sample_single.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Create object of sample_single class — sample_single","text":"named list class sample_single. contains following: ids vector characters containing ids subjects included original dataset. beta numeric vector estimated regression coefficients. sigma list estimated covariance matrices (one level vars$group). theta numeric vector transformed covariances. failed logical. TRUE model fit failed. ids_samp vector characters containing ids subjects included given sample.","code":""},{"path":"/reference/scalerConstructor.html","id":null,"dir":"Reference","previous_headings":"","what":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"Scales design matrix non-categorical columns mean 0 standard deviation 1.","code":""},{"path":"/reference/scalerConstructor.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"object initialisation used determine relevant mean SD's scale scaling (un-scaling) performed relevant object methods. Un-scaling done linear model Beta Sigma coefficients. purpose first column dataset scaled assumed outcome variable variables assumed post-transformation predictor variables (.e. dummy variables already expanded).","code":""},{"path":"/reference/scalerConstructor.html","id":"public-fields","dir":"Reference","previous_headings":"","what":"Public fields","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"centre Vector column means. first value outcome variable, variables predictors. scales Vector column standard deviations. first value outcome variable, variables predictors.","code":""},{"path":[]},{"path":"/reference/scalerConstructor.html","id":"public-methods","dir":"Reference","previous_headings":"","what":"Public methods","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"scalerConstructor$new() scalerConstructor$scale() scalerConstructor$unscale_sigma() scalerConstructor$unscale_beta() scalerConstructor$clone()","code":""},{"path":"/reference/scalerConstructor.html","id":"method-new-","dir":"Reference","previous_headings":"","what":"Method new()","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"Uses dat determine relevant column means standard deviations use scaling un-scaling future datasets. Implicitly assumes new datasets column order dat","code":""},{"path":"/reference/scalerConstructor.html","id":"usage","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"","code":"scalerConstructor$new(dat)"},{"path":"/reference/scalerConstructor.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"dat data.frame matrix. columns must numeric (.e dummy variables, must already expanded ).","code":""},{"path":"/reference/scalerConstructor.html","id":"details-1","dir":"Reference","previous_headings":"","what":"Details","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"Categorical columns (determined values entirely 1 0) scaled. achieved setting corresponding values centre 0 scale 1.","code":""},{"path":"/reference/scalerConstructor.html","id":"method-scale-","dir":"Reference","previous_headings":"","what":"Method scale()","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"Scales dataset continuous variables mean 0 standard deviation 1.","code":""},{"path":"/reference/scalerConstructor.html","id":"usage-1","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"","code":"scalerConstructor$scale(dat)"},{"path":"/reference/scalerConstructor.html","id":"arguments-1","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"dat data.frame matrix whose columns numeric (.e. dummy variables expanded ) whose columns order dataset used initialization function.","code":""},{"path":"/reference/scalerConstructor.html","id":"method-unscale-sigma-","dir":"Reference","previous_headings":"","what":"Method unscale_sigma()","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"Unscales sigma value (matrix) estimated linear model using design matrix scaled object. function works first column initialisation data.frame outcome variable.","code":""},{"path":"/reference/scalerConstructor.html","id":"usage-2","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"","code":"scalerConstructor$unscale_sigma(sigma)"},{"path":"/reference/scalerConstructor.html","id":"arguments-2","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"sigma numeric value matrix.","code":""},{"path":"/reference/scalerConstructor.html","id":"returns","dir":"Reference","previous_headings":"","what":"Returns","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"numeric value matrix","code":""},{"path":"/reference/scalerConstructor.html","id":"method-unscale-beta-","dir":"Reference","previous_headings":"","what":"Method unscale_beta()","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"Unscales beta value (vector) estimated linear model using design matrix scaled object. function works first column initialization data.frame outcome variable.","code":""},{"path":"/reference/scalerConstructor.html","id":"usage-3","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"","code":"scalerConstructor$unscale_beta(beta)"},{"path":"/reference/scalerConstructor.html","id":"arguments-3","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"beta numeric vector beta coefficients estimated linear model.","code":""},{"path":"/reference/scalerConstructor.html","id":"returns-1","dir":"Reference","previous_headings":"","what":"Returns","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"numeric vector.","code":""},{"path":"/reference/scalerConstructor.html","id":"method-clone-","dir":"Reference","previous_headings":"","what":"Method clone()","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"objects class cloneable method.","code":""},{"path":"/reference/scalerConstructor.html","id":"usage-4","dir":"Reference","previous_headings":"","what":"Usage","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"","code":"scalerConstructor$clone(deep = FALSE)"},{"path":"/reference/scalerConstructor.html","id":"arguments-4","dir":"Reference","previous_headings":"","what":"Arguments","title":"R6 Class for scaling (and un-scaling) design matrices — scalerConstructor","text":"deep Whether make deep clone.","code":""},{"path":"/reference/set_simul_pars.html","id":null,"dir":"Reference","previous_headings":"","what":"Set simulation parameters of a study group. — set_simul_pars","title":"Set simulation parameters of a study group. — set_simul_pars","text":"function provides input arguments study group needed simulate data simulate_data(). simulate_data() generates data two-arms clinical trial longitudinal continuous outcomes two intercurrent events (ICEs). ICE1 may thought discontinuation study treatment due study drug condition related (SDCR) reasons. ICE2 may thought discontinuation study treatment due uninformative study drop-, .e. due study drug condition related (NSDRC) reasons outcome data ICE2 always missing.","code":""},{"path":"/reference/set_simul_pars.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Set simulation parameters of a study group. — set_simul_pars","text":"","code":"set_simul_pars( mu, sigma, n, prob_ice1 = 0, or_outcome_ice1 = 1, prob_post_ice1_dropout = 0, prob_ice2 = 0, prob_miss = 0 )"},{"path":"/reference/set_simul_pars.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Set simulation parameters of a study group. — set_simul_pars","text":"mu Numeric vector describing mean outcome trajectory visit (including baseline) assuming ICEs. sigma Covariance matrix outcome trajectory assuming ICEs. n Number subjects belonging group. prob_ice1 Numeric vector specifies probability experiencing ICE1 (discontinuation study treatment due SDCR reasons) visit subject observed outcome visit equal mean baseline (mu[1]). single numeric provided, probability applied visit. or_outcome_ice1 Numeric value specifies odds ratio experiencing ICE1 visit corresponding +1 higher value observed outcome visit. prob_post_ice1_dropout Numeric value specifies probability study drop-following ICE1. subject simulated drop-ICE1, outcomes ICE1 set missing. prob_ice2 Numeric specifies additional probability post-baseline visit affected study drop-. Outcome data subject's first simulated visit affected study drop-subsequent visits set missing. generates second intercurrent event ICE2, may thought treatment discontinuation due NSDRC reasons subsequent drop-. subject, ICE1 ICE2 simulated occur, assumed earlier counts. case ICEs simulated occur time, assumed ICE1 counts. means single subject can experience either ICE1 ICE2, . prob_miss Numeric value specifies additional probability given post-baseline observation missing. can used produce \"intermittent\" missing values associated ICE.","code":""},{"path":"/reference/set_simul_pars.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Set simulation parameters of a study group. — set_simul_pars","text":"simul_pars object named list containing simulation parameters.","code":""},{"path":"/reference/set_simul_pars.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Set simulation parameters of a study group. — set_simul_pars","text":"details, please see simulate_data().","code":""},{"path":[]},{"path":"/reference/set_vars.html","id":null,"dir":"Reference","previous_headings":"","what":"Set key variables — set_vars","title":"Set key variables — set_vars","text":"function used define names key variables within data.frame's provided input arguments draws() ancova().","code":""},{"path":"/reference/set_vars.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Set key variables — set_vars","text":"","code":"set_vars( subjid = \"subjid\", visit = \"visit\", outcome = \"outcome\", group = \"group\", covariates = character(0), strata = group, strategy = \"strategy\" )"},{"path":"/reference/set_vars.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Set key variables — set_vars","text":"subjid name \"Subject ID\" variable. length 1 character vector. visit name \"Visit\" variable. length 1 character vector. outcome name \"Outcome\" variable. length 1 character vector. group name \"Group\" variable. length 1 character vector. covariates name covariates used context modeling. See details. strata name stratification variable used context bootstrap sampling. See details. strategy name \"strategy\" variable. length 1 character vector.","code":""},{"path":"/reference/set_vars.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Set key variables — set_vars","text":"draws() ancova() covariates argument can specified indicate variables included imputation analysis models respectively. wish include interaction terms need manually specified .e. covariates = c(\"group*visit\", \"age*sex\"). Please note use () function inhibit interpretation/conversion objects supported. Currently strata used draws() combination method_condmean(type = \"bootstrap\") method_approxbayes() order allow specification stratified bootstrap sampling. default strata set equal value group assumed users want preserve group size samples. See draws() details. Likewise, currently strategy argument used draws() specify name strategy variable within data_ice data.frame. See draws() details.","code":""},{"path":[]},{"path":"/reference/set_vars.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Set key variables — set_vars","text":"","code":"if (FALSE) { # Using CDISC variable names as an example set_vars( subjid = \"usubjid\", visit = \"avisit\", outcome = \"aval\", group = \"arm\", covariates = c(\"bwt\", \"bht\", \"arm * avisit\"), strategy = \"strat\" ) }"},{"path":"/reference/simulate_data.html","id":null,"dir":"Reference","previous_headings":"","what":"Generate data — simulate_data","title":"Generate data — simulate_data","text":"Generate data two-arms clinical trial longitudinal continuous outcome two intercurrent events (ICEs). ICE1 may thought discontinuation study treatment due study drug condition related (SDCR) reasons. ICE2 may thought discontinuation study treatment due uninformative study drop-, .e. due study drug condition related (NSDRC) reasons outcome data ICE2 always missing.","code":""},{"path":"/reference/simulate_data.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Generate data — simulate_data","text":"","code":"simulate_data(pars_c, pars_t, post_ice1_traj, strategies = getStrategies())"},{"path":"/reference/simulate_data.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Generate data — simulate_data","text":"pars_c simul_pars object generated set_simul_pars(). specifies simulation parameters control arm. pars_t simul_pars object generated set_simul_pars(). specifies simulation parameters treatment arm. post_ice1_traj string specifies observed outcomes occurring ICE1 simulated. Must target function included strategies. Possible choices : Missing Random \"MAR\", Jump Reference \"JR\", Copy Reference \"CR\", Copy Increments Reference \"CIR\", Last Mean Carried Forward \"LMCF\". User-defined strategies also added. See getStrategies() details. strategies named list functions. Default equal getStrategies(). See getStrategies() details.","code":""},{"path":"/reference/simulate_data.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Generate data — simulate_data","text":"data.frame containing simulated data. includes following variables: id: Factor variable specifies id subject. visit: Factor variable specifies visit assessment. Visit 0 denotes baseline visit. group: Factor variable specifies treatment group subject belongs . outcome_bl: Numeric variable specifies baseline outcome. outcome_noICE: Numeric variable specifies longitudinal outcome assuming ICEs. ind_ice1: Binary variable takes value 1 corresponding visit affected ICE1 0 otherwise. dropout_ice1: Binary variable takes value 1 corresponding visit affected drop-following ICE1 0 otherwise. ind_ice2: Binary variable takes value 1 corresponding visit affected ICE2. outcome: Numeric variable specifies longitudinal outcome including ICE1, ICE2 intermittent missing values.","code":""},{"path":"/reference/simulate_data.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Generate data — simulate_data","text":"data generation works follows: Generate outcome data visits (including baseline) multivariate normal distribution parameters pars_c$mu pars_c$sigma control arm parameters pars_t$mu pars_t$sigma treatment arm, respectively. Note randomized trial, outcomes distribution baseline treatment groups, .e. one set pars_c$mu[1]=pars_t$mu[1] pars_c$sigma[1,1]=pars_t$sigma[1,1]. Simulate whether ICE1 (study treatment discontinuation due SDCR reasons) occurs visit according parameters pars_c$prob_ice1 pars_c$or_outcome_ice1 control arm pars_t$prob_ice1 pars_t$or_outcome_ice1 treatment arm, respectively. Simulate drop-following ICE1 according pars_c$prob_post_ice1_dropout pars_t$prob_post_ice1_dropout. Simulate additional uninformative study drop-probabilities pars_c$prob_ice2 pars_t$prob_ice2 visit. generates second intercurrent event ICE2, may thought treatment discontinuation due NSDRC reasons subsequent drop-. simulated time drop-subject's first visit affected drop-data visit subsequent visits consequently set missing. subject, ICE1 ICE2 simulated occur, assumed earlier counts. case ICEs simulated occur time, assumed ICE1 counts. means single subject can experience either ICE1 ICE2, . Adjust trajectories ICE1 according given assumption expressed post_ice1_traj argument. Note post-ICE1 outcomes intervention arm can adjusted. Post-ICE1 outcomes control arm adjusted. Simulate additional intermittent missing outcome data per arguments pars_c$prob_miss pars_t$prob_miss. probability ICE visit modeled according following logistic regression model: ~ 1 + (visit == 0) + ... + (visit == n_visits-1) + ((x-alpha)) : n_visits number visits (including baseline). alpha baseline outcome mean. term ((x-alpha)) specifies dependency probability ICE current outcome value. corresponding regression coefficients logistic model defined follows: intercept set 0, coefficients corresponding discontinuation visit subject outcome equal mean baseline set according parameters pars_c$prob_ice1 (pars_t$prob_ice1), regression coefficient associated covariate ((x-alpha)) set log(pars_c$or_outcome_ice1) (log(pars_t$or_outcome_ice1)). Please note baseline outcome missing affected ICEs.","code":""},{"path":"/reference/simulate_dropout.html","id":null,"dir":"Reference","previous_headings":"","what":"Simulate drop-out — simulate_dropout","title":"Simulate drop-out — simulate_dropout","text":"Simulate drop-","code":""},{"path":"/reference/simulate_dropout.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Simulate drop-out — simulate_dropout","text":"","code":"simulate_dropout(prob_dropout, ids, subset = rep(1, length(ids)))"},{"path":"/reference/simulate_dropout.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Simulate drop-out — simulate_dropout","text":"prob_dropout Numeric specifies probability post-baseline visit affected study drop-. ids Factor variable specifies id subject. subset Binary variable specifies subset affected drop-. .e. subset binary vector length equal length ids takes value 1 corresponding visit affected drop-0 otherwise.","code":""},{"path":"/reference/simulate_dropout.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Simulate drop-out — simulate_dropout","text":"binary vector length equal length ids takes value 1 corresponding outcome affected study drop-.","code":""},{"path":"/reference/simulate_dropout.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Simulate drop-out — simulate_dropout","text":"subset can used specify outcome values affected drop-. default subset set 1 values except values corresponding baseline outcome, since baseline supposed affected drop-. Even subset specified user, values corresponding baseline outcome still hard-coded 0.","code":""},{"path":"/reference/simulate_ice.html","id":null,"dir":"Reference","previous_headings":"","what":"Simulate intercurrent event — simulate_ice","title":"Simulate intercurrent event — simulate_ice","text":"Simulate intercurrent event","code":""},{"path":"/reference/simulate_ice.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Simulate intercurrent event — simulate_ice","text":"","code":"simulate_ice(outcome, visits, ids, prob_ice, or_outcome_ice, baseline_mean)"},{"path":"/reference/simulate_ice.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Simulate intercurrent event — simulate_ice","text":"outcome Numeric variable specifies longitudinal outcome single group. visits Factor variable specifies visit assessment. ids Factor variable specifies id subject. prob_ice Numeric vector specifies visit probability experiencing ICE current visit subject outcome equal mean baseline. single numeric provided, probability applied visit. or_outcome_ice Numeric value specifies odds ratio ICE corresponding +1 higher value outcome visit. baseline_mean Mean outcome value baseline.","code":""},{"path":"/reference/simulate_ice.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Simulate intercurrent event — simulate_ice","text":"binary variable takes value 1 corresponding outcome affected ICE 0 otherwise.","code":""},{"path":"/reference/simulate_ice.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Simulate intercurrent event — simulate_ice","text":"probability ICE visit modeled according following logistic regression model: ~ 1 + (visit == 0) + ... + (visit == n_visits-1) + ((x-alpha)) : n_visits number visits (including baseline). alpha baseline outcome mean set via argument baseline_mean. term ((x-alpha)) specifies dependency probability ICE current outcome value. corresponding regression coefficients logistic model defined follows: intercept set 0, coefficients corresponding discontinuation visit subject outcome equal mean baseline set according parameter or_outcome_ice, regression coefficient associated covariate ((x-alpha)) set log(or_outcome_ice).","code":""},{"path":"/reference/simulate_test_data.html","id":null,"dir":"Reference","previous_headings":"","what":"Create simulated datasets — simulate_test_data","title":"Create simulated datasets — simulate_test_data","text":"Creates longitudinal dataset format rbmi designed analyse.","code":""},{"path":"/reference/simulate_test_data.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Create simulated datasets — simulate_test_data","text":"","code":"simulate_test_data( n = 200, sd = c(3, 5, 7), cor = c(0.1, 0.7, 0.4), mu = list(int = 10, age = 3, sex = 2, trt = c(0, 4, 8), visit = c(0, 1, 2)) ) as_vcov(sd, cor)"},{"path":"/reference/simulate_test_data.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Create simulated datasets — simulate_test_data","text":"n number subjects sample. Total number observations returned thus n * length(sd) sd standard deviations outcome visit. .e. square root diagonal covariance matrix outcome cor correlation coefficients outcome values visit. See details. mu coefficients use construct mean outcome value visit. Must named list elements int, age, sex, trt & visit. See details.","code":""},{"path":"/reference/simulate_test_data.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Create simulated datasets — simulate_test_data","text":"number visits determined size variance covariance matrix. .e. 3 standard deviation values provided 3 visits per patient created. covariates simulated dataset produced follows: Patients age sampled random N(0,1) distribution Patients sex sampled random 50/50 split Patients group sampled random fixed group n/2 patients outcome variable sampled multivariate normal distribution, see details mean outcome variable derived : coefficients intercept, age sex taken mu$int, mu$age mu$sex respectively, must length 1 numeric. Treatment visit coefficients taken mu$trt mu$visit respectively must either length 1 (.e. constant affect across visits) equal number visits (determined length sd). .e. wanted treatment slope 5 visit slope 1 specify: correlation matrix constructed cor follows. Let cor = c(, b, c, d, e, f) correlation matrix :","code":"outcome = Intercept + age + sex + visit + treatment mu = list(..., \"trt\" = c(0,5,10), \"visit\" = c(0,1,2)) 1 a b d a 1 c e b c 1 f d e f 1"},{"path":"/reference/sort_by.html","id":null,"dir":"Reference","previous_headings":"","what":"Sort data.frame — sort_by","title":"Sort data.frame — sort_by","text":"Sorts data.frame (ascending default) based upon variables within dataset","code":""},{"path":"/reference/sort_by.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Sort data.frame — sort_by","text":"","code":"sort_by(df, vars = NULL, decreasing = FALSE)"},{"path":"/reference/sort_by.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Sort data.frame — sort_by","text":"df data.frame vars character vector variables decreasing logical whether sort order descending ascending (default) order. Can either single logical value (case applied variables) vector length vars","code":""},{"path":"/reference/sort_by.html","id":"ref-examples","dir":"Reference","previous_headings":"","what":"Examples","title":"Sort data.frame — sort_by","text":"","code":"if (FALSE) { sort_by(iris, c(\"Sepal.Length\", \"Sepal.Width\"), decreasing = c(TRUE, FALSE)) }"},{"path":"/reference/split_dim.html","id":null,"dir":"Reference","previous_headings":"","what":"Transform array into list of arrays — split_dim","title":"Transform array into list of arrays — split_dim","text":"Transform array list arrays listing performed given dimension.","code":""},{"path":"/reference/split_dim.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Transform array into list of arrays — split_dim","text":"","code":"split_dim(a, n)"},{"path":"/reference/split_dim.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Transform array into list of arrays — split_dim","text":"Array number dimensions least 2. n Positive integer. Dimension listed.","code":""},{"path":"/reference/split_dim.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Transform array into list of arrays — split_dim","text":"list length n arrays number dimensions equal number dimensions minus 1.","code":""},{"path":"/reference/split_dim.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Transform array into list of arrays — split_dim","text":"example, 3 dimensional array n = 1, split_dim(,n) returns list 2 dimensional arrays (.e. list matrices) element list [, , ], takes values 1 length first dimension array. Example: inputs: <- array( c(1,2,3,4,5,6,7,8,9,10,11,12), dim = c(3,2,2)), means : n <- 1 output res <- split_dim(,n) list 3 elements:","code":"a[1,,] a[2,,] a[3,,] [,1] [,2] [,1] [,2] [,1] [,2] --------- --------- --------- 1 7 2 8 3 9 4 10 5 11 6 12 res[[1]] res[[2]] res[[3]] [,1] [,2] [,1] [,2] [,1] [,2] --------- --------- --------- 1 7 2 8 3 9 4 10 5 11 6 12"},{"path":"/reference/split_imputations.html","id":null,"dir":"Reference","previous_headings":"","what":"Split a flat list of imputation_single() into multiple imputation_df()'s by ID — split_imputations","title":"Split a flat list of imputation_single() into multiple imputation_df()'s by ID — split_imputations","text":"Split flat list imputation_single() multiple imputation_df()'s ID","code":""},{"path":"/reference/split_imputations.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Split a flat list of imputation_single() into multiple imputation_df()'s by ID — split_imputations","text":"","code":"split_imputations(list_of_singles, split_ids)"},{"path":"/reference/split_imputations.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Split a flat list of imputation_single() into multiple imputation_df()'s by ID — split_imputations","text":"list_of_singles list imputation_single()'s split_ids list 1 element per required split. element must contain vector \"ID\"'s correspond imputation_single() ID's required within sample. total number ID's must equal length list_of_singles","code":""},{"path":"/reference/split_imputations.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Split a flat list of imputation_single() into multiple imputation_df()'s by ID — split_imputations","text":"function converts list imputations structured per patient structured per sample .e. converts :","code":"obj <- list( imputation_single(\"Ben\", numeric(0)), imputation_single(\"Ben\", numeric(0)), imputation_single(\"Ben\", numeric(0)), imputation_single(\"Harry\", c(1, 2)), imputation_single(\"Phil\", c(3, 4)), imputation_single(\"Phil\", c(5, 6)), imputation_single(\"Tom\", c(7, 8, 9)) ) index <- list( c(\"Ben\", \"Harry\", \"Phil\", \"Tom\"), c(\"Ben\", \"Ben\", \"Phil\") ) output <- list( imputation_df( imputation_single(id = \"Ben\", values = numeric(0)), imputation_single(id = \"Harry\", values = c(1, 2)), imputation_single(id = \"Phil\", values = c(3, 4)), imputation_single(id = \"Tom\", values = c(7, 8, 9)) ), imputation_df( imputation_single(id = \"Ben\", values = numeric(0)), imputation_single(id = \"Ben\", values = numeric(0)), imputation_single(id = \"Phil\", values = c(5, 6)) ) )"},{"path":"/reference/str_contains.html","id":null,"dir":"Reference","previous_headings":"","what":"Does a string contain a substring — str_contains","title":"Does a string contain a substring — str_contains","text":"Returns vector TRUE/FALSE element x contains element subs .e.","code":"str_contains( c(\"ben\", \"tom\", \"harry\"), c(\"e\", \"y\")) [1] TRUE FALSE TRUE"},{"path":"/reference/str_contains.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Does a string contain a substring — str_contains","text":"","code":"str_contains(x, subs)"},{"path":"/reference/str_contains.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Does a string contain a substring — str_contains","text":"x character vector subs character vector substrings look ","code":""},{"path":"/reference/strategies.html","id":null,"dir":"Reference","previous_headings":"","what":"Strategies — strategies","title":"Strategies — strategies","text":"functions used implement various reference based imputation strategies combining subjects distribution reference distribution based upon visits failed meet Missing--Random (MAR) assumption.","code":""},{"path":"/reference/strategies.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Strategies — strategies","text":"","code":"strategy_MAR(pars_group, pars_ref, index_mar) strategy_JR(pars_group, pars_ref, index_mar) strategy_CR(pars_group, pars_ref, index_mar) strategy_CIR(pars_group, pars_ref, index_mar) strategy_LMCF(pars_group, pars_ref, index_mar)"},{"path":"/reference/strategies.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Strategies — strategies","text":"pars_group list parameters subject's group. See details. pars_ref list parameters subject's reference group. See details. index_mar logical vector indicating visits meet MAR assumption subject. .e. identifies observations non-MAR intercurrent event (ICE).","code":""},{"path":"/reference/strategies.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Strategies — strategies","text":"pars_group pars_ref must list containing elements mu sigma. mu must numeric vector sigma must square matrix symmetric covariance matrix dimensions equal length mu index_mar. e.g. Users can define strategy functions include via strategies argument impute() using getStrategies(). said following strategies available \"box\": Missing Random (MAR) Jump Reference (JR) Copy Reference (CR) Copy Increments Reference (CIR) Last Mean Carried Forward (LMCF)","code":"list( mu = c(1,2,3), sigma = matrix(c(4,3,2,3,5,4,2,4,6), nrow = 3, ncol = 3) )"},{"path":"/reference/string_pad.html","id":null,"dir":"Reference","previous_headings":"","what":"string_pad — string_pad","title":"string_pad — string_pad","text":"Utility function used replicate str_pad. Adds white space either end string get equal desired length","code":""},{"path":"/reference/string_pad.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"string_pad — string_pad","text":"","code":"string_pad(x, width)"},{"path":"/reference/string_pad.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"string_pad — string_pad","text":"x string width desired length","code":""},{"path":"/reference/transpose_imputations.html","id":null,"dir":"Reference","previous_headings":"","what":"Transpose imputations — transpose_imputations","title":"Transpose imputations — transpose_imputations","text":"Takes imputation_df object transposes e.g.","code":"list( list(id = \"a\", values = c(1,2,3)), list(id = \"b\", values = c(4,5,6) ) )"},{"path":"/reference/transpose_imputations.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Transpose imputations — transpose_imputations","text":"","code":"transpose_imputations(imputations)"},{"path":"/reference/transpose_imputations.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Transpose imputations — transpose_imputations","text":"imputations imputation_df object created imputation_df()","code":""},{"path":"/reference/transpose_imputations.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Transpose imputations — transpose_imputations","text":"becomes","code":"list( ids = c(\"a\", \"b\"), values = c(1,2,3,4,5,6) )"},{"path":"/reference/transpose_results.html","id":null,"dir":"Reference","previous_headings":"","what":"Transpose results object — transpose_results","title":"Transpose results object — transpose_results","text":"Transposes Results object (created analyse()) order group estimates together vectors.","code":""},{"path":"/reference/transpose_results.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Transpose results object — transpose_results","text":"","code":"transpose_results(results, components)"},{"path":"/reference/transpose_results.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Transpose results object — transpose_results","text":"results list results. components character vector components extract (.e. \"est\", \"se\").","code":""},{"path":"/reference/transpose_results.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Transpose results object — transpose_results","text":"Essentially function takes object format: produces:","code":"x <- list( list( \"trt1\" = list( est = 1, se = 2 ), \"trt2\" = list( est = 3, se = 4 ) ), list( \"trt1\" = list( est = 5, se = 6 ), \"trt2\" = list( est = 7, se = 8 ) ) ) list( trt1 = list( est = c(1,5), se = c(2,6) ), trt2 = list( est = c(3,7), se = c(4,8) ) )"},{"path":"/reference/transpose_samples.html","id":null,"dir":"Reference","previous_headings":"","what":"Transpose samples — transpose_samples","title":"Transpose samples — transpose_samples","text":"Transposes samples generated draws() grouped subjid instead sample number.","code":""},{"path":"/reference/transpose_samples.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Transpose samples — transpose_samples","text":"","code":"transpose_samples(samples)"},{"path":"/reference/transpose_samples.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Transpose samples — transpose_samples","text":"samples list samples generated draws().","code":""},{"path":"/reference/validate.analysis.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate analysis objects — validate.analysis","title":"Validate analysis objects — validate.analysis","text":"Validates return object analyse() function.","code":""},{"path":"/reference/validate.analysis.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate analysis objects — validate.analysis","text":"","code":"# S3 method for analysis validate(x, ...)"},{"path":"/reference/validate.analysis.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate analysis objects — validate.analysis","text":"x analysis results object (class \"jackknife\", \"bootstrap\", \"rubin\"). ... used.","code":""},{"path":"/reference/validate.draws.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate draws object — validate.draws","title":"Validate draws object — validate.draws","text":"Validate draws object","code":""},{"path":"/reference/validate.draws.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate draws object — validate.draws","text":"","code":"# S3 method for draws validate(x, ...)"},{"path":"/reference/validate.draws.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate draws object — validate.draws","text":"x draws object generated as_draws(). ... used.","code":""},{"path":"/reference/validate.html","id":null,"dir":"Reference","previous_headings":"","what":"Generic validation method — validate","title":"Generic validation method — validate","text":"function used perform assertions object conforms expected structure basic assumptions violated. throw error checks pass.","code":""},{"path":"/reference/validate.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Generic validation method — validate","text":"","code":"validate(x, ...)"},{"path":"/reference/validate.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Generic validation method — validate","text":"x object validated. ... additional arguments pass specific validation method.","code":""},{"path":"/reference/validate.is_mar.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate is_mar for a given subject — validate.is_mar","title":"Validate is_mar for a given subject — validate.is_mar","text":"Checks longitudinal data patient divided MAR followed non-MAR data; non-MAR observation followed MAR observation allowed.","code":""},{"path":"/reference/validate.is_mar.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate is_mar for a given subject — validate.is_mar","text":"","code":"# S3 method for is_mar validate(x, ...)"},{"path":"/reference/validate.is_mar.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate is_mar for a given subject — validate.is_mar","text":"x Object class is_mar. Logical vector indicating whether observations MAR. ... used.","code":""},{"path":"/reference/validate.is_mar.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Validate is_mar for a given subject — validate.is_mar","text":"error issue otherwise return TRUE.","code":""},{"path":"/reference/validate.ivars.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate inputs for vars — validate.ivars","title":"Validate inputs for vars — validate.ivars","text":"Checks required variable names defined within vars appropriate datatypes","code":""},{"path":"/reference/validate.ivars.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate inputs for vars — validate.ivars","text":"","code":"# S3 method for ivars validate(x, ...)"},{"path":"/reference/validate.ivars.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate inputs for vars — validate.ivars","text":"x named list indicating names key variables source dataset ... used","code":""},{"path":"/reference/validate.references.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate user supplied references — validate.references","title":"Validate user supplied references — validate.references","text":"Checks ensure user specified references expect values (.e. found within source data).","code":""},{"path":"/reference/validate.references.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate user supplied references — validate.references","text":"","code":"# S3 method for references validate(x, control, ...)"},{"path":"/reference/validate.references.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate user supplied references — validate.references","text":"x named character vector. control factor variable (group variable source dataset). ... used.","code":""},{"path":"/reference/validate.references.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Validate user supplied references — validate.references","text":"error issue otherwise return TRUE.","code":""},{"path":"/reference/validate.sample_list.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate sample_list object — validate.sample_list","title":"Validate sample_list object — validate.sample_list","text":"Validate sample_list object","code":""},{"path":"/reference/validate.sample_list.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate sample_list object — validate.sample_list","text":"","code":"# S3 method for sample_list validate(x, ...)"},{"path":"/reference/validate.sample_list.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate sample_list object — validate.sample_list","text":"x sample_list object generated sample_list(). ... used.","code":""},{"path":"/reference/validate.sample_single.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate sample_single object — validate.sample_single","title":"Validate sample_single object — validate.sample_single","text":"Validate sample_single object","code":""},{"path":"/reference/validate.sample_single.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate sample_single object — validate.sample_single","text":"","code":"# S3 method for sample_single validate(x, ...)"},{"path":"/reference/validate.sample_single.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate sample_single object — validate.sample_single","text":"x sample_single object generated sample_single(). ... used.","code":""},{"path":"/reference/validate.simul_pars.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate a simul_pars object — validate.simul_pars","title":"Validate a simul_pars object — validate.simul_pars","text":"Validate simul_pars object","code":""},{"path":"/reference/validate.simul_pars.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate a simul_pars object — validate.simul_pars","text":"","code":"# S3 method for simul_pars validate(x, ...)"},{"path":"/reference/validate.simul_pars.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate a simul_pars object — validate.simul_pars","text":"x simul_pars object generated set_simul_pars(). ... used.","code":""},{"path":"/reference/validate.stan_data.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate a stan_data object — validate.stan_data","title":"Validate a stan_data object — validate.stan_data","text":"Validate stan_data object","code":""},{"path":"/reference/validate.stan_data.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate a stan_data object — validate.stan_data","text":"","code":"# S3 method for stan_data validate(x, ...)"},{"path":"/reference/validate.stan_data.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate a stan_data object — validate.stan_data","text":"x stan_data object. ... used.","code":""},{"path":"/reference/validate_analyse_pars.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate analysis results — validate_analyse_pars","title":"Validate analysis results — validate_analyse_pars","text":"Validates analysis results generated analyse().","code":""},{"path":"/reference/validate_analyse_pars.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate analysis results — validate_analyse_pars","text":"","code":"validate_analyse_pars(results, pars)"},{"path":"/reference/validate_analyse_pars.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate analysis results — validate_analyse_pars","text":"results list results generated analysis fun used analyse(). pars list expected parameters analysis. lists .e. c(\"est\", \"se\", \"df\").","code":""},{"path":"/reference/validate_datalong.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate a longdata object — validate_datalong","title":"Validate a longdata object — validate_datalong","text":"Validate longdata object","code":""},{"path":"/reference/validate_datalong.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate a longdata object — validate_datalong","text":"","code":"validate_datalong(data, vars) validate_datalong_varExists(data, vars) validate_datalong_types(data, vars) validate_datalong_notMissing(data, vars) validate_datalong_complete(data, vars) validate_datalong_unifromStrata(data, vars) validate_dataice(data, data_ice, vars, update = FALSE)"},{"path":"/reference/validate_datalong.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate a longdata object — validate_datalong","text":"data data.frame containing longitudinal outcome data + covariates multiple subjects vars vars object created set_vars() data_ice data.frame containing subjects ICE data. See draws() details. update logical, indicates ICE data set first time update applied","code":""},{"path":"/reference/validate_datalong.html","id":"details","dir":"Reference","previous_headings":"","what":"Details","title":"Validate a longdata object — validate_datalong","text":"functions used validate various different parts longdata object used draws(), impute(), analyse() pool(). particular: validate_datalong_varExists - Checks variable listed vars actually exists data validate_datalong_types - Checks types key variable expected .e. visit factor variable validate_datalong_notMissing - Checks none key variables (except outcome variable) contain missing values validate_datalong_complete - Checks data complete .e. 1 row subject * visit combination. e.g. nrow(data) == length(unique(subjects)) * length(unique(visits)) validate_datalong_unifromStrata - Checks make sure variables listed stratification variables vary time. e.g. subjects switch stratification groups.","code":""},{"path":"/reference/validate_strategies.html","id":null,"dir":"Reference","previous_headings":"","what":"Validate user specified strategies — validate_strategies","title":"Validate user specified strategies — validate_strategies","text":"Compares user provided strategies required (reference). throw error values reference defined.","code":""},{"path":"/reference/validate_strategies.html","id":"ref-usage","dir":"Reference","previous_headings":"","what":"Usage","title":"Validate user specified strategies — validate_strategies","text":"","code":"validate_strategies(strategies, reference)"},{"path":"/reference/validate_strategies.html","id":"arguments","dir":"Reference","previous_headings":"","what":"Arguments","title":"Validate user specified strategies — validate_strategies","text":"strategies named list strategies. reference list character vector strategies need defined.","code":""},{"path":"/reference/validate_strategies.html","id":"value","dir":"Reference","previous_headings":"","what":"Value","title":"Validate user specified strategies — validate_strategies","text":"throw error issue otherwise return TRUE.","code":""},{"path":"/news/index.html","id":"rbmi-development-version","dir":"Changelog","previous_headings":"","what":"rbmi (development version)","title":"rbmi (development version)","text":"Include vignette obtain frequentist information-anchored inference conditional mean imputation using rbmi Added FAQ vignette","code":""},{"path":"/news/index.html","id":"rbmi-126","dir":"Changelog","previous_headings":"","what":"rbmi 1.2.6","title":"rbmi 1.2.6","text":"CRAN release: 2023-11-24 Updated unit tests fix false-positive error CRAN’s testing servers","code":""},{"path":"/news/index.html","id":"rbmi-125","dir":"Changelog","previous_headings":"","what":"rbmi 1.2.5","title":"rbmi 1.2.5","text":"CRAN release: 2023-09-20 Updated internal Stan code ensure future compatibility (@andrjohns, #390) Updated package description include relevant references (#393) Fixed documentation typos (#393)","code":""},{"path":"/news/index.html","id":"rbmi-123","dir":"Changelog","previous_headings":"","what":"rbmi 1.2.3","title":"rbmi 1.2.3","text":"CRAN release: 2022-11-14 Minor internal tweaks ensure compatibility packages rbmi depends ","code":""},{"path":"/news/index.html","id":"rbmi-121","dir":"Changelog","previous_headings":"","what":"rbmi 1.2.1","title":"rbmi 1.2.1","text":"CRAN release: 2022-10-25 Removed native pipes |> testing code package backwards compatible older servers Replaced glmmTMB dependency mmrm package. resulted package stable (less model fitting convergence issues) well speeding run times 3-fold.","code":""},{"path":"/news/index.html","id":"rbmi-114","dir":"Changelog","previous_headings":"","what":"rbmi 1.1.4","title":"rbmi 1.1.4","text":"CRAN release: 2022-05-18 Updated urls references vignettes Fixed bug visit factor levels re-constructed incorrectly delta_template() Fixed bug wrong visit displayed error message specific visit doesn’t data draws() Fixed bug wrong input parameter displayed error message simulate_data()","code":""},{"path":"/news/index.html","id":"rbmi-111--113","dir":"Changelog","previous_headings":"","what":"rbmi 1.1.1 & 1.1.3","title":"rbmi 1.1.1 & 1.1.3","text":"CRAN release: 2022-03-08 change functionality 1.1.0 Various minor tweaks address CRAN checks messages","code":""},{"path":"/news/index.html","id":"rbmi-110","dir":"Changelog","previous_headings":"","what":"rbmi 1.1.0","title":"rbmi 1.1.0","text":"CRAN release: 2022-03-02 Initial public release","code":""}] diff --git a/pkgdown.yml b/pkgdown.yml index e462e476..245aa40c 100644 --- a/pkgdown.yml +++ b/pkgdown.yml @@ -1,9 +1,11 @@ -pandoc: 2.17.1.1 -pkgdown: 2.0.3 +pandoc: 3.1.1 +pkgdown: 2.0.7 pkgdown_sha: ~ articles: + CondMean_Inference: CondMean_Inference.html + FAQ: FAQ.html advanced: advanced.html quickstart: quickstart.html stat_specs: stat_specs.html -last_built: 2022-06-15T07:45Z +last_built: 2024-03-15T16:04Z