On the reliability of Bayesian posterior clade probabilities in phylogenetic analysis

Abstract

This article discusses possible reasons why posterior clade probabilities obtained from Bayesian phylogenetic analyses might be inaccurate. It attempts to list all possible sources of uncertainty and error in Bayesian phylogenetic analysis. The choice of priors on trees has been suggested by several authors as a cause of inaccurate posterior clade probabilities. I argue strongly for using priors based on biological knowledge. Two possible sources of bias which do not seem to have been published before are also discussed. One is a computational problem and the other related to majority rule consensus trees.