Choice of topology estimators in Bayesian phylogenetic analysis.

Wheeler WC and Pickett KM (2008. Topology-Bayes versus clade-Bayes in phylogenetic analysis. Mol Biol Evol. 25:447-453.) discuss two ways of summarizing the posterior probability distribution of a Bayesian phylogenetic analysis, which they refer to as "topology-Bayes" and "clade-Bayes." They claim that the clade-Bayes approach leads to problems such as "exaggerated clade support, inconsistently biased priors, and the impossibility of topology hypothesis testing," which are not problems for the topology-Bayes approach. However, their argument for topology-Bayes over clade-Bayes is based on errors in the interpretation of summary statistics associated with Bayesian phylogenetic analysis. Although there is a well-documented difference between the maximum posterior probability topology and the majority-rule consensus topology (the established terms for topology-Bayes and clade-Bayes summaries, respectively), both have a place in phylogenetic analysis. Choice of summarization strategy should be driven by choice of parameters that need to be estimated versus those to be marginalized given the evolutionary questions being asked or hypotheses being tested.