Table 2 Integrating over multiple models for stochastic character mapping.

Three transition rate models (ARD, ER, and IR) and one mixed model were considered for stochastic character mapping. The number of simulations (Nsim) for each model was normalized using Akaike information criterion weights (wAIC) for that model and the mixed model sampled character histories from all three models. The transition rate matrix for a given model was either fixed at its empirical MLE or sampled using MCMC from its posterior distribution (MCMC). Rates (q), AIC scores, Likelihood (LnL), mean number of forward and reverse transitions, and PP of red blood reconstructed for the most recent common ancestor (MRCA) for all Prasinohaema are shown for the ONE-p70 tree. Models are described in the main text.

Transition rate matrixModelModel summaryMean no. of changesPP of red-blooded MRCA
qRGqGRAICwAICln (L)NsimRed→GreenRed←Green
MLE (fixed)ARD11.7137.232.20.386−14.13864.09.027.1
ER11.811.832.50.332−15.33324.20.599.9
IR10.60.032.80.282−15.42824.00.0100
Mixed10003.92.281.4
MCMCARD3.51.036.30.156−16.11563.90.1100
ER3.83.833.50.622−15.76223.90.1100
IR3.90.035.50.222−16.72224.00.0100
Mixed10003.90.1100