For Bayesian tree reconstructions in MrBayes, we conducted four Markov Chain Monte Carlo (MCMC) runs with a default temperature of 0.2 and the TrN + I + G model as selected as best-fit model in jModeltest 2.1 [ 72] under the Bayesian information criterion (BIC) and the Decision Theory Performance-based Selection (DT).
Figure 6 shows the AUC for each model in the stepwise selection procedure with the best model, as selected using the Akaike Information Criterion, identified.
Table 4 AIC and coefficients of discrimination (D) values of the different forms of the model fitted to the dataset using logistic regression AIC D Adaptation Belief 484.86 0.01 Adaptation Belief + Experience 482.12 0.03 Full model 196.29 0.65 The value in bold is the most parsimonious model as selected by AIC comparison (ΔAIC > 2).
Both ML and BI analyses were conducted under a GTR+I+Γ model as selected by the AIC and a single partition.
We used the GTR+I+G model as selected by MrModeltest for the combined dataset and ran analyses until 10,000 generations revealed no significant improvement of the likelihood scores of the topology.
The GTR + IG substitution model, as selected by jModelTest, and a Yule tree prior were used.
For MrBayes, we used the LG + G + I model, as selected by ProtTest.
Because MrBayes 3.1.2 only implements 1, 2, and 6 substitution rate models, it was often not possible to implement the preferred model as selected by the AIC.
Because MrBayes 3.1.2 only implements 1, 2, and 6 substitutions rate models, often it was not possible to implement the preferred model as selected by the AIC.
Our ML analysis served as the guide tree, and the ML statistical method was chosen using the GTR model, as selected by jModelTest2.
