This repository contains the software implementations for PRP. The software calculates the (prior/posterior)-replication predictive p-value, which is a statistical assessment of replicability via bayesian moodel criticism.

The repository includes source R code and scripts to replicate all the simulation results described in the manuscript. The data for real application is included in the R package.

Source code for R package PRP is included in R_src. To install, run

devtools::install_github("ArtemisZhao/PRP/R_src")

The code and data for simulation and real data analysis are included in the folder PRP_paper. They should enable readers to fully reproduce the simulation results.

To reproduce simulation and analysis results, clone the repo and run make in each sub-directory. The simulation data will be re-generated.

The data for the real application described in the paper are included in the R package.

library(PRP)
## RP:P 
data("RPP_filtered")

## Cardiovascular disease impact on the COVID-19 mortality
data("mortality")

## Cardiovascular disease impact on the COVID-19 severity
data("severity")

The following R packages are required to install and run the analysis code:

• devtools
• mvtnorm
• metafor
• dplyr
• ggplot2

A docker image with pre-configured Linux running environment and pre-installed R libraries is availabel for download from docker hub.

• Yi Zhao (zhayi at umich dot edu)
• Xiaoquan Wen (xwen at umich dot edu)

• Zhao, Y. and Wen, X. Statistical Assessment of Replicability via Bayesian Model Criticism. arXiv:2105.03993