Gitlab repository/sandbox related to subgroup shrinkage simulation study (ongoing work not final).
To run the benchmark pipeline you need to have a recent R installed. Please also install the following R packages
- cmdstan & cmdstanr
- targets, tarchetypes, ps, qs2
- tidyverse
- brms, posterior, bayesplot, tidybayes
- RBesT
- assertthat
- survival
- scoringRules
The model benchmark is controlled via targets for each data set. The sources come with a normal and a TTE endpoint fake data simulation example. For each data set the respective targets pipeline can be started using
make
to run the two pipelines in parallel, one can use make -j2
It is strongly recommended to configure a high-performance cluster or
other means of parallelization within the _targets_setup.R
configuration file in order to speed up the pipeline.