Tail Paradox, Partial Identifiability, and Influential Priors in Bayesian Branch Length Inference
Open accessThe DOI matches its Crossref record; 5 of the 5 references it cites were found in a registry or an index.
Abstract
Recent studies have observed that Bayesian analyses of sequence data sets using the program MrBayes sometimes generate extremely large branch lengths, with posterior credibility intervals for the tree length (sum of branch lengths) excluding the maximum likelihood estimates. Suggested explanations for this phenomenon include the existence of multiple local peaks in the posterior, lack of convergence of the chain in the tail of the posterior, mixing problems, and misspecified priors on branch lengths. Here, we analyze the behavior of Bayesian Markov chain Monte Carlo algorithms when the chain is in the tail of the posterior distribution and note that all these phenomena can occur. In Bayesian phylogenetics, the likelihood function approaches a constant instead of zero when the branch lengths increase to infinity. The flat tail of the likelihood can cause poor mixing and undue influence of the prior. We suggest that the main cause of the extreme branch length estimates produced in many Bayesian analyses is the poor choice of a default prior on branch lengths in current Bayesian phylogenetic programs. The default prior in MrBayes assigns independent and identical distributions to branch lengths, imposing strong (and unreasonable) assumptions about the tree length. The problem is exacerbated by the strong correlation between the branch lengths and parameters in models of variable rates among sites or among site partitions. To resolve the problem, we suggest two multivariate priors for the branch lengths (called compound Dirichlet priors) that are fairly diffuse and demonstrate their utility in the special case of branch length estimation on a star phylogeny. Our analysis highlights the need for careful thought in the specification of high-dimensional priors in Bayesian analyses.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Relaxed Phylogenetics and Dating with Confidence
Drummond, A. J., Ho, S. Y. W., Phillips, M. J., & Rambaut, A. (2006) · PLoS Biology · doi:10.1371/journal.pbio.0040088 · Cited by 6,463 · cited 2×
“Another promising approach is to use relaxed-clock models to infer both rooted and unrooted trees (Drummond et al. 2006), with the prior for branch lengths generated by specifying a prior on the divergence times and another prior on the rates for branches based on a model of stochastic rate drift…”
BackgroundYang, Z. (1994) · Journal of Molecular Evolution · doi:10.1007/bf00160154 · Cited by 2,960 · cited 1×
“We analyze two data sets (one real and one simulated) to explore the correlation between the shape parameter α of the gamma distribution for variable rates among sites (Yang 1994) and the branch lengths.”
MethodSelf-citationEstimating the rate of evolution of the rate of molecular evolution
Thorne, J. L., Kishino, H., & Painter, I. S. (1998) · Molecular Biology and Evolution · doi:10.1093/oxfordjournals.molbev.a025892 · Cited by 1,317 · cited 3×
“We follow the suggestion of Yang and Rannala (2005) to have different prior means for the s − 3 internal branch lengths and the s external branch lengths.”
BackgroundBayesian Selection of Continuous-Time Markov Chain Evolutionary Models
Suchard, M. A., Weiss, R. E., & Sinsheimer, J. S. (2001) · Molecular Biology and Evolution · doi:10.1093/oxfordjournals.molbev.a003872 · Cited by 791 · cited 2×
“(18) However, an inverse-gamma hyperprior is assigned on μ: f (μ|α,β) = β α Γ(α) μ−α−1e−β/μ. (19) Suchard et al. (2001) recommended α = 2.1 and β = 1.1 but such a value of α is extreme and the prior mean branch length of 1.0 seems rather large.”
BackgroundInfluentialBranch-length prior influences Bayesian posterior probability of phylogeny.
Yang, Z., & Rannala, B. (2005) · Systematic Biology · doi:10.1080/10635150590945313 · Cited by 228 · cited 2×
“This was suggested as a way of reducing the high posterior probabilities for trees and clades, similar to assigning nonzero probabilities tomultifurcating trees (Lewis et al. 2005).”
BackgroundSelf-citation
Cited by
Molecular phylogenetics: principles and practice
Yang, Z., & Rannala, B. (2012) · Nature Reviews Genetics · doi:10.1038/nrg3186 · Cited by 752
Self-citationThe Influence of Gene Flow on Species Tree Estimation: A Simulation Study
Leaché, A. D., Harris, R. B., Rannala, B., & Yang, Z. (2014) · Systematic Biology · doi:10.1093/sysbio/syt049 · Cited by 368
Self-citationMolecular Evolution: A Statistical Approach
Yang, Z. (2014) · doi:10.1093/acprof:oso/9780199602605.001.0001 · Cited by 273
Self-citationA biologist’s guide to Bayesian phylogenetic analysis
Nascimento, F. F., Reis, M. D., & Yang, Z. (2017) · Nature Ecology & Evolution · doi:10.1038/s41559-017-0280-x · Cited by 212
Self-citationThe impact of the rate prior on Bayesian estimation of divergence times with multiple Loci
Dos Reis, M., Zhu, T., & Yang, Z. (2014) · Systematic Biology · doi:10.1093/sysbio/syu020 · Cited by 93
Self-citation
Metadata
- Authors
- Rannala, B., Zhu, T., & Yang, Z.
- Venue
- Molecular Biology and Evolution
- Publisher
- Oxford University Press (OUP)
- Published
- 2011-09-02
- Volume
- 29
- Issue
- 1
- Pages
- 325-335
- ISSN
- 0737-4038 (ISSN-L)
- 1537-1719
- OpenAlex
- W2125562015
- Cited by
- 113
- References
- 5
- FWCI
- 4.70
- Type
- journal article
- Language
- en
- Open access
- bronze
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.