Normalized power priors always discount historical data

This repository contains code and data related to the paper

Pawel, S., Aust, F., Held, L., and Wagenmakers, E.-J. (2023). Normalized power priors always discount historical data. Stat, 12( 1), e591. DOI:10.1002/sta4.591

Reproducing the results

We offer two ways to reproduce the results

1. Reproduction with local computational environment (requires R and LaTeX)

First install the required R packages by running in a shell from the root directory of the repository

R -e 'install.packages(read.delim("CRANpackages.txt", header = FALSE)[,1])'

Then run

cd paper
make pdf

this should reproduce all analyses and output the file pppooling.pdf in the paper directory.

Although our analysis depends on only few dependencies, this approach may lead to different results (or not even compile successfully) in the future if R or an R package dependency changes. The R and R package versions which were used in our analysis can be seen in the output of the sessionInfo command at the bottom of the manuscript in the snapshot of the GitHub repository at the time of submission.

2. Reproduction within Docker container (requires Docker with root rights)

Run in a shell from the root directory of the repository

make drunpdf

this should output the file pppooling.pdf in the paper directory. The Docker approach takes a bit longer but reruns our analyses in a Docker container which encapsulates the computational environment (R and R package versions) that was used in the original analysis.