While I was off being really busy, an interesting project to learn PyMC was discussed on their mailing list, beginning thusly:
I am trying to learn PyMC and I decided to start from the very simple discrete Sprinkler model.
I have found plenty of examples for continuous models, but I am not sure how should I proceed with conditional tables, especially when the condition is over more than a variable. How would you model a CPT with PyMC, and in particular the Sprinkler model?
I spend all my time on continuous models these days, and I miss my old topics from back in grad school. Not that this is one of them, exactly. But when I found myself with a long wait while running some validation code, I thought I’d give it a try. The model turned out to be simple, although using MCMC to fit it is probably not the best idea. Continue reading