This github repo was meant to primarily be a performance comparison between MCMC sampling implementations between pymc3 and pystan. Additionally, it served as a learning tool as I tried to replicate the "classic" examples found most commonly online. The plan is to implement three different models: 1) Hierarchical Regression; 2) Switchpoints; and 3) Time-Series Model. As of 20180902 only Hierarchical Regression is implemented.
After cloning this repo, navigate to the repo location and set up the conda virtual environment referencing the "bayesvs.txt" file by:
source setup_env.bashThe configuration file is a .py file that holds key model information and configuration parameters. For the radon model, there is a flag on whether you want to download an updated radon file from Gelman's website: http://www.stat.columbia.edu/~gelman/arm/examples/radon_complete/srrs2.dat
The configuration file also controls the number of iterations used in the MCMC samplers, the number of parallel jobs where applicable, as well as the stan model code.