We investigated Dirichlet and BDS priors because different specifications of these priors may be suited to particular node age distributions, for example, when most cladogenesis events are ancient and clustered near the root.
The probability density function of the Dirichlet is unimodal, symmetrical, and located centrally when the parameter α .
The joint prior density of divergence times, conditional on root age, and other time constraints are derived from the BDS model.
Consider a tree with The program BEAST implements special cases of the BDS model to generate a time prior (Drummond and Rambaut 2007).
One hundred single locus data sets were generated for each of the following sequence lengths: 250, 500, 1000, 5000, and 10,000 bp for each of the 4 trees.
The 4 trees that were used to simulate sequence data. The simulated data were analyzed with soft or hard minimal and maximal constraints of 8.1 and 9.9 time units, respectively, specified for the roots of all trees.
One obvious difference between these phylogenies and those with older root ages is that taxa are less divergent, and so sequences tend to be less informative. This appears to be the pattern seen in some real data sets (Brown et al. The prior may therefore have a greater impact in analyses of shallow phylogenies. Internode times along paths from internal nodes to the tips are then obtained as proportions of the oldest node age on the path (which is the root node on the first iteration) from a Dirichlet density.
Divergence time estimation may be less accurate and/or precise as a result. In this article, we focus on 2 different methods for specifying the prior on divergence times. Generation of internode times is repeated on different paths within the tree until all nodes have been assigned ages.Different divergence times priors within MCMCTREE, MULTIDIVTIME, and BEAST were used to analyze mitochondrial DNA data sets from a Bovid subfamily (the Caprinae) from Asian lizards.Posterior divergence times were quite sensitive to different BDS priors but less sensitive to different Dirichlet priors.This is implemented in the program MCMCTREE (Rannala and Yang 2007; Yang and Rannala 2006; Yang 2007).The BDS prior is specified by 3 parameters: per-lineage birth (λ) and death (μ) rates and the sampling proportion (ρ).This typically occurred when the 95% prior interval did not include the true age and/or sequence lengths were ≤ 1 kbp.