Hello. I want to run an MCMC analysis on BayesTraits using a set of Bayesian trees I created using MrBayes. I want to run the analysis using both the consensus tree and a sample of 1,000 trees to check the effect of using the consensus tree VS the sample of trees. My MrBayes run produced ~35,000 trees.
First of all, am I allowed to use a subset of these trees, let's say 1,000?
Secondly, should I take a random sample from the ~35,000 trees, or should I get the first 1,000 with the highest posterior probability?
Many thanks in advance!
Hello Ioannis, my inclination is to take a random sample of the 35,000 trees. That, in my opinion, would be the best way to account for error if you didn't want to use them all (using all may be computationally intensive). If those 35,000 trees are post-burn-in and the MCMC reached convergence, then all of those trees should merely be samples around the likelihood peak that the MCMC converged on (i.e., they should be quite similar to, and within error of, one another). Moreover, according to the BayesTraits manual, the MCMC samples trees in proportion to their likelihood.
Thank you Jacob. That's what I did and I also did the analysis using all the trees and I got similar results.
- Badges
- Science method
- Similar topics
- Biological Science
- Evolutionary Biology
- Phylogenetics
- Phylogenetic Analysis