As for our phylogenetic analysis, we used PartitionFinder v1.1.0
(Lanfear et al., 2012) to select the best model of nucleotide substitution
and the best partitioning scheme for our 37-taxon dataset
after defining nine potential partitions (3 coding regions 3 codon
positions). The best scheme again combined the data for the three
genes in two partitions (codon positions 1 + 2 and codon position
3), with TrN+I+G identified as the best model of substitution for
both partitions. Thus, for our BEAST runs we used the same parameter
settings for each partition – Substitution Model = TN93; Base
Frequencies = Estimated; Site Heterogeneity Model = Gamma
(with 4 rate categories); Clock Model = Lognormal relaxed clock
(uncorrelated) – and we allowed evolutionary rates and base
frequencies to be unlinked between partitions. We assumed a Speciation: Yule Process model for the Tree Prior (which is generally
appropriate for trees involving sequences from several species),
used a randomly-generated starting tree (subject to the
monophyly constraints noted above), and specified a diffuse
gamma prior distribution (with shape 0.001 and scale 1000) for
the ucld.mean parameter.