Table 1.

Likelihood and AIC Values under Various Partitioning Schemes

Partitioning	No. of Partitions	Parameters (k)	ln(L)	AIC	ΔAIC	2 × ln ΔBF	RBF
None	1	9	−1,279,328.877	2,558,675.754	105,496.41	21.76	0.059
Forward/Reverse	2	18	−1,258,902.112	2,517,840.225	64,660.88	20.79	0.058
Homogeneous/Heterogeneous	2	18	−1,273,139.835	2,546,315.669	93,136.33	21.51	0.060
Gene	14	126	−1,256,482.92	2,513,217.84	60,038.51	20.64	0.082
Codon 1 + 2 + 3	3	27	−1,251,864.871	2,503,783.742	50,604.41	20.30	0.058
Codon 1 + 2 + 3 + Forward/Reverse	6	54	−1,229,360.303	2,458,828.606	5,649.26	16.11	0.050
Gene × codon	42	378	−1,226,211.669	2,453,179.339	n/a	n/a	n/a

Note.—The likelihood of the data under each partitioning scheme was assessed on the fixed topology of a randomized parsimony tree under a GTR + G model, with the number of partitions, free parameters, and ln(L) scores used in the calculations given. ΔAIC refers to the decrease in likelihood relative to the most complex model (partitioning by gene and codon). Values for 2 × ln ΔBF₁₀ > 10 are usually considered to be highly significant. RBF was calculated according to Castoe et al. (2005) as 2 × ln ΔBF10/Δ parameters, to penalize greater model complexity.