Skip to main content

Table 2 The models evaluated in the simulation-based analyses

From: An improved approximate-Bayesian model-choice method for estimating shared evolutionary history

 

Priors

Model

t

Ï„

θ

 

M m s B a y e s

t∼D U{1,…,Y}

τ∼U(0,10 [ 25 M G A])

θ A ∼U(0,0.05)

θ ̄ D ∼U(0,0.05)

M U s h a p e d

t∼D U{1,…,Y}

τ∼E x p(m e a n=2.887[ 7.22M G A])

θ A ∼θD1∼θD2∼E x p(m e a n=0.025)

M U n i f o r m

t∼D U{a(Y)}

τ∼E x p(m e a n=2.887[ 7.22M G A])

θ A ∼θD1∼θD2∼E x p(m e a n=0.025)

M D P P

t∼D P(χ∼G a m m a(·,·))

τ∼E x p(m e a n=2.887[ 7.22M G A])

θ A ∼θD1∼θD2∼E x p(m e a n=0.025)

  1. For the M D P P model, the prior on the concentration parameter, χ∼ Gamma (·,·), was set to Gamma(2,2) for the validation analyses and Gamma(1.5,18.1) for the power analyses. The distributions of divergence times are given in units of 4N C generations followed in brackets by units of millions of generations ago (MGA), with the former converted to the latter assuming a per-site rate of 1×10−8 mutations per generation. For model M m s B a y e s , the priors for theta parameters are θ A ∼ U(0, 0.05) and θD1,θD2∼ Beta(1, 1)×2× U(0, 0.05). The later is summarized as θ ̄ D ∼ U(0, 0.05). For the M D P P and M U n i f o r m , and M U s h a p e d models, θ A ,θD1, and θD2 are independently and exponentially distributed with a mean of 0.025.