Data partitioning in Bayesian analysis: molecular phylogenetics of metalmark moths (Lepidoptera: Choreutidae)

In this study a multilocus phylogenetic analysis of metalmark moths (Lepidoptera: Choreutidae) focused on resolving the higher‐level phylogeny of this group is presented. Through the analysis of this dataset, I explore different data‐partitioning strategies in Bayesian phylogenetic inference, and find that a partitioning strategy can have a large influence on the results of phylogenetic analysis. Depending on how the data are partitioned, there can be significant differences in branch support. I also test for the existence of the Bayesian star tree paradox, and its importance in this dataset, and find that it appears to inflate support for the clade including Rhobonda gaurisana, Hemerophila houttuinialis, H. diva and H. felis, but plays no role in other cases where the differences between maximum‐likelihood bootstraps and Bayesian posterior probabilities are large. The results of all the phylogenetic analyses strongly suggest that including Millieriinae in Choreutidae renders the family polyphyletic. The monophyly of the other two subfamilies, Brenthiinae and Choreutinae, as well as their sister‐group relationship, is strongly supported. Similarly, the monophyly of all the genera examined except Hemerophila is also well supported. To bring the classification of Choreutidae in line with our current understanding of the phylogenetic relationships in the family, I propose to exclude Millieriinae from Choreutidae, elevate it to Millieriidae Heppner, and place it as incertae sedis within Ditrysia.