MCMC and the guts of JAGS

In this section we look at different aspects of how JAGS works. You may want to treat it as reference material to come back to from time to time when working through the other units.

Generating posterior draws

This is an important topic, but there is already quite a lot on the web, in particular at mmeredith.net. So for the moment we will give some links and promise to come back later and do a nice page!

A simple Gibbs sampler

MCMC samples with a random walk

MCMC settings

Even if you are happy to treat the generation of draws as a “black box”, it’s still helpful to understand the workflow of JAGS as this illuminates the meaning of the settings you put in when running a model – or the defaults which you use.

MCMC diagnostics

Once you have the output with draws from the posterior distribution, it’s important to check the output for bias and precision.

Leave a Reply

Your email address will not be published. Required fields are marked *