Mildew

This is the supplementary R code for the paper

[Jousimo J, Tack AJM, Ovaskainen O, Mononen T, Susi H, Tollenaere C, Laine A-L. Ecological and evolutionary effects of fragmentation on infectious disease dynamics. Science 344 (6189): 1289-1293] (http://www.sciencemag.org/content/344/6189/1289.abstract).

to preprocess the data, estimate the spatio-temporal models for each response and plot the results.

Setup

Install R-INLA testing version with the R command

source("http://www.math.ntnu.no/inla/givemeINLA-testing.R")

Then run Mildew setup script with the command

source_url("https://raw.github.com/statguy/R-Mildew/master/inst/process/setup.R")

Be sure you have the devtools package installed first.

Optional: For parallel processing on a HPC, there is a Python script. However, it needs to be configured for your local system first.

Code files

In the subdirectory process of the Mildew package installation directory (which is shown by the command path.package("Mildew") after loading the Mildew package), you can find the following files to preprocess, estimate and report results for the mildew data:

• preprocess.R contains high-level code to preprocess the data so that it is useful for the analysis.
• estimate.R contains high-level code to estimate all models.
• reports.R contains high-level code to load all results and print reports.

Configuration

Set basePath in to point your mildew data directory, for example

basePath <- "~/mildew"

Preprocessing, estimation, reporting

You may run the code from command line at the installation root directory with

R --vanilla --args 1 < inst/process/preprocess.R
R --vanilla --args 11 < inst/process/estimate.R
R --vanilla < inst/process/reports.R

where the number is a task id. For preprocess, there are three tasks, one for each response. Task id 1 must be run first as the results are used for the other responses. For estimate, there are 6 * 3 tasks, so each response has six models and the task ids run as 11, 12, 13, 21, 22, 23,..., 61, 62, 63.

With the current configuration, estimation may take up to 1 day per model depending on your setup. You may want to construct a mesh with less nodes by adjusting the mesh parameters occ.mesh.params, col.mesh.params, ext.mesh.params in estimate.R or use a subset of the data for testing. Use the plotMesh method, e.g.

occ$plotMesh()

to plot the mesh.

Feedback

Send any feedback to [email protected].