Table 2.

Akaike Information Criterion (AIC) sensitivity analyses about the effect of model parameters on log-likelihood gains provided by codon partitions.

	Model of DNA evolution
	No codon partition	BC effect	GTR effect	Γ effect	BL effect	BC + GTR + Γ + BL effect
CYB	23,021.0	22,742.4	22,954.8	22,509.1	22,689.3	21,901.9
GHR	8,352.0	8,306.1	8,347.7	8,346.2	8,434.3	8,388.3

In the simplest model – without codon partition–, a single set of base composition (BC), substitution rate (GTR), among-sites rate heterogeneity (Γ), and branch length (BL) parameters is used, irrespective of the CYB and GHR codon positions. In the most complex model, one independent set of BC, GTR, Γ, and BL parameters is attributed to each codon position of either CYB or GHR. For intermediate models, the effect of giving a set of either BC, GTR, Γ, or BL parameters to each codon partition is measured through the AIC (= -2 × log-likelihood of the reference topology [Figure 2] + 2 × number of free parameters). The best, i.e., lowest, AIC are in bold for each gene.