How Trustworthy Is Your Tree? Bayesian Phylogenetic Effective Sample Size Through the Lens of Monte Carlo Error

Bayesian inference is a popular and widely-used approach to infer phylogenies (evolutionary trees). However, despite decades of widespread application, it remains difficult judge how well given Markov chain Monte Carlo (MCMC) run explores the space phylogenetic trees. In this paper, we investigate error phylogenies, focusing on high-dimensional summaries posterior distribution, including variability in estimated edge/branch (known phylogenetics as “split”) probabilities tree probabilities, summary tree. Specifically, ask if there any measure effective sample size (ESS) applicable trees which capable capturing these three measures. We find that are some ESS measures inherent using MCMC samples approximate distributions phylogenies. term measures, identify set useful practice for assessing error. Lastly, present visualization tools can improve comparisons between multiple independent runs by accounting each chain. Our results indicate common post-MCMC workflows insufficient capture tree, highlight need both within-chain mixing between-chain convergence assessments.