Stan

Robert Myles McDonnell

4 minute read

(This turned out to be a bit of a ramble, for the code go here 😄)

My PhD thesis focused on latent variable models as a way to model legislative voting behaviour. The main model I used is called the Bayesian Item Response model, and the idea is that, from the observed votes of the legislators, we can build a scale on which we can place them, relative to one another.

5 minute read

Anybody who has ever tried to run even a moderately-sized Bayesian IRT model in R (for ideal points as in the political science literature, or otherwise) will know that these models can take a long time. It’s not R’s fault: these are usually big models with lots of parameters, and naturally take longer.1 Not to mention the fact that Bayesian computation is more computationally intense than other methods. Historically (okay, I’m talking about the last twenty years, maybe ‘historically’ is a little strong), the sampling software BUGS (Bayesian Inference Using Gibbs Sampling) and then JAGS were used to run Bayesian models (JAGS is still pretty common, and BUGS too, though not as much).