Research compendium for our article:

Repository Overview

• data contains raw data used to fit our apixaban model as well as derived datasets from our modelling pricess. The data/README provides a codebook for the data in each file.

• analysis contains the scripts that fit the models, compute appropriate metrics, and produce figures. The analysis/README outlines what datasets and what figures are produced by which scripts.

• figures contains figures generated for the paper.

• models contains relevant Stan models used in the paper. The models/README includes a summary of the models and in which scripts they are used.

The shell script build.sh will delete all figures and then rerun the scripts in analysis in order to generate them again. Fitting our simulation via Hamiltonian Monte Carlo can take up to an hour on a 2017 iMac with 8GB of RAM.

The branch cmdstanr implements the study using cmdstanr rather than rstan. If you are using OSX Catalina, there have been some notorious problems getting rstan to sample in parallel. cmdstanr is one way to get around this.