EmpPrior: using outside empirical data to inform branch-length priors for Bayesian phylogenetics

BMC Bioinformatics

Table 1 Default and informed tree-length estimates

Outside Dataset	λ	α	c	MLE	95 % HPD
None (default)	10			0.40	[10.54–14.94]
		1	1	0.40	[0.24–0.34]
S1777	95.6			0.40	[0.83–1.25]
		0.12	1	0.40	[0.30–0.48]
		1	0.18	0.40	[0.26–0.39]
		0.11	1.35	0.40	[0.30–0.49]
S2043	80.7			0.40	[1.03–1.53]
		0.13	1	0.40	[0.30–0.49]
		1	0.27	0.40	[0.26–0.38]
		0.10	1.93	0.40	[0.29–0.46]

Highest posterior density intervals (95 % HPDs) of tree length (TL) for analyses of a focal dataset from brittle stars [12]. Each individual row corresponds to a Bayesian analysis of the focal data with different prior settings for branch- and tree-lengths. We used both default and informed parameterizations of the exponential (λ) and compound Dirichlet (α, c) branch-length priors. The top set of two shaded rows show default settings and focal inferences. Below that, alternate shadings indicate prior settings parameterized with different outside datasets (S1777 and S2043). The focal TL MLE was produced with Garli [13]. Focal HPDs in bold contain the MLE

ISSN: 1471-2105