/in-clinic-updates-PSA

Code for fast updates of patient-level variables in hierarchical Bayesian models, incorporating new, out-of-sample data.

Primary LanguageR

Prostate Surveillance

Prediction modeling for active surveillance of prostate cancer

Analysts can reproduce our results using a sun grid engine computing cluster, by running

bash run-MCMC-IS.sh

This will ultimately result in plots being saved to the plots folder.

Contents

Folders

  • simulation-data - contains code for simulation data based on the Johns Hopkins Active Surveillance cohort (JHAS), and results of these simulations.
  • plots - folder to save final plots of the analysis

Files

  • run-MCMC-IS.sh - run the pipeline for MCMC and for IS
  • call-jags-cluster.R - run MCMC using jags. This function calls the script
    • call-jags-functions-setup.R, which in turn calls either
      • model-for-jags-IOP_BX-IOP_SURG.txt,
      • model-for-jags-NIOP_BX-IOP_SURG.txt,
      • model-for-jags-IOP_BX-NIOP_SURG.txt, or
      • model-for-jags-NIOP_BX-NIOP_SURG.txt, depending on variables that tell whether informative observation processes for biopsies and surgeries should be included in the model.
  • combine-jags-results.R - concatenates the results of parallel runs of call-jags-cluster.R
  • IS-fitting.R - run IS on individual patients. This function calls the script
    • IS-function-setup.R
  • IS-combine-results.r - aggregate results from IS-fitting.R and generate plots