Friday, April 22, 2011

MCMC method: more on analyzing the posterior sample

I've been working a little more on analyzing the posterior sample from our Bayesian MCMC method for identifying the location of a single shift in the evolutionary rate for a continuously valued character. I originally described this method (which has also been submitted for publication with several coauthors) on the blog here. I'm not done with this, but I created an animation of 801 samples from the posterior distribution created by a 100,000 generation MCMC run on the simulated dataset and tree given here. In the animation, the branches of the tree are stretched or shortened depending on their relative rates for that sample. You can see that most of the posterior density for the rate shift is on the generating edge, but that some samples occasionally stray to different parts of the phylogenetic tree. I will be updating the code for all of these analyses very soon.
I have to admit that this animation is not nearly as neat my animation of all bi- and multifurcating trees (although at least it is shorter). I recommend watching it with Robert Randolph "Jesus is Just Alright" playing in the background (as I did). It will surely seem cooler that way.

No comments:

Post a Comment