Generate posterior expectations / predictions for new data

[stefgehrig]

Is there a simple way to create posterior predictions / posterior expectations of predictions from a model object and the fitted model after sampling from the posterior (i.e., after running something like engine.sample_all_epochs())? The intention would be to supply a new data frame, similar as is possible in R with many model types via predict(model, newdata = …).

In {brms}, for example, there is such functionality via brms::posterior_predict() and brms::posterior_epred(), respectively.

I'm not a firm Python user, but this would probably be best done within Python?

[jobrachem]

I'm afraid that, currently, Liesel does not offer much convenience functionality for predictions. You get the posterior samples, but have to implement custom functionality for predictions. We have an issue for that here: #54 But unfortunately I don't expect that we will have the time to prioritise prediction convenience in the near future.

[stefgehrig]

Makes total sense, it is not a must-have. Thank you!