Jake Vanderplas’s comparison of Python MCMC modules was preceded by a Bayesian polemic. In general, I find the stats philosophy war old-timey and distracting, but his comparison of confidence intervals and credible intervals is something I need to understand better.
Tag Archives: MCMC
While I’m on a microblogging spree, I’ve been meaning to link to this informative comparison of pymc, emcee, and pystan: http://jakevdp.github.io/blog/2014/06/14/frequentism-and-bayesianism-4-bayesian-in-python/
Here is an interesting StackOverflow question, about a model for data distributed as the sum of two uniforms with unknown support. I was surprised how hard it was for me.
I think the future of probabilistic programming should be to make a model for this easy to code.
I’ve had no teaching responsibilities over the last quarter, and I must miss it. I’ve found myself responding to PyMC questions on StackOverflow more than ever before. It is an interesting window into what is hard in Bayesian computation. Checking (and achieving) MCMC convergence is one thing that is hard. Here are some questions and answers that include it:
The PyMC documentation is a little slim on the topic of defining a custom sampler, and I had to figure it out for some DisMod work over the years. Here is a minimal example of how I did it, in answer to a CrossValidated question.
I have revisited my approach to deciding if MCMC has run for long enough recently, and I’m collecting some of the relevant material here:
Last time I thought about it: http://healthyalgorithms.com/2010/04/19/practical-mcmc-advice-when-to-stop/
Original paper for R_hat approach: http://www.stat.columbia.edu/~gelman/research/published/itsim.pdf
Presentation comparing several approaches: http://www.people.fas.harvard.edu/~plam/teaching/methods/convergence/convergence_print.pdf
Published comparison: http://www.jstor.org/stable/2291683
Blog about a cool visual approach: http://andrewgelman.com/2009/12/24/visualizations_1/
Discussion on cross-validated: http://stats.stackexchange.com/questions/507/what-is-the-best-method-for-checking-convergence-in-mcmc
Book with a chapter on this referenced there: http://www.amazon.com/dp/1441915753/?tag=stackoverfl08-20 (available as an eBook from UW Library, how convenient!)
Another related blog: http://xianblog.wordpress.com/2012/11/28/mcmc-convergence-assessment/
A recent question on the PyMC mailing list inspired me to make a really inefficient version of the Naive Bayes classifier. Enjoy.
There has been a low murmur about new MCMC package bouncing through my email inbox for a while now. Stan, it is. The project has reached the point where the developers are soliciting Python integration volunteers, so I decided it is time to check it out.
Good news, it installed and ran the example without frustration! I don’t take that for granted with research software.