Why some clades have low bootstrap frequencies and high Bayesian posterior probabilities

Abstract

Bayesian posterior probabilities are wrongly considered by many systematists as indicative of character support, and equivalent to non-parametric bootstrap frequencies. Here I argue against this view. Non-parametric bootstrap is indicative of the amount of evidence in a data matrix supporting each clade in the tree, while Bayesian posterior probabilities are not intended to represent that property. Clades with high posterior probability may not have a large amount of characters favouring them, and their frequencies are the result of the particular sampling procedure of the Bayesian Markov chain Monte Carlo method, which tends to sample very similar topologies according to their posterior probabilities. Both metrics may relate to the notion of confidence, but depict different properties.

Independent host switching events by digenean parasites of cetaceans inferred from ribosomal DNA

“…The clade P. gastrophilus + C. formosanus was supported by a high PP (100%), indicating that this clade occurred in all the topologies sampled in the Bayesian MCMC. However, the ML bootstrap value was only 59%, indicating that there is not enough character support for this clade (see García-Sandoval, 2014). In the SSU and ITS2 rDNA analysis, P. gastrophilus was placed together with two species of the Heterophyidae, i.e.…”

Estimation of 12S marker fragment effectiveness for ancient phylogeny reconstruction

“…Results: two different methods of phylogeny reconstruction -maximum likelihood and Bayesian Inference -yielded overall similar but slightly different topologies. the bootstrap node supports produced by maximum likelihood reconstruction were generally lower than Bayesian posterior supports which may be due to a loss of information during bootstrap resampling [4]. However both methods showed a significant decrease of node support values from tips to root of the trees.…”

Opuntia delafuentiana (Cactaceae: Opuntioideae), a new xoconostle from central Mexico

“…Once the standard deviation between the chains was stabilized at a value of 0.01, the sample size for the parameters were examined with Tracer 1.6 (Rambaut et al 2014, beast.bio.ed.ac.uk/Tracer). The results were summarized in a majority rule consensus computed with MrBayes 3.2, and posterior probabilities and character support values were drawn over the maximum clade credibility tree recovered with TreeAnotator 2.1.2 (Bouckaert et al 2014, beast2.org), as recommended by García-Sandoval (2014).…”