I expect many people have their own MCMC (Markov chain Mote Carlo) implementations for Bayesian parameter estimation (in multiple programming languages) and indeed I also have my own. However, rather than keeping my Matlab implementation to myself I’ve decided to release it as “yamm” (Yet Another Matlab MCMC code) on github. This is obviously not the only Matlab MCMC code (see e.g. here or here for a couple of examples), and I make no claim for it being the best optimised, fastest or most efficient code (it’s very much at best beta in terms of a release), but hopefully it might prove useful to others.
In writing the code I acknowledge various code that has helped it’s development. The various proposal distributions were helped greatly by the work of Veitch & Vecchio (arXiv:0911.3820) and implementations currently in the LALinference software suite. This includes the affine invariant ensemble samplers of Goodman & Weare as also implemented in the python emcee software by Foreman-Mackey et al. (arXiv:1202.3665).
Any comments or suggestions about the software are welcome here. This code is quite similar to the Matlab Nested Sampler code described here.