Investigating outgroup taxon sampling: empirical examples from Noctuoidea (Lepidoptera) and bees (Hymenoptera: Apoidea)

While taxon sampling has been investigated thoroughly on many simulated data sets, an astonishing lack of attention has been paid to the taxa selected as outgroups, especially in real data sets. To investigate this problem, two data sets (featuring Noctuoidea as an ingroup and closely related families as outgroup taxa; and featuring bees as an ingroup and apoid wasps and closely related vespoid wasps as outgroup taxa) were chosen where a large number of outgroup taxa were available. A random sampling approach was applied to each group, featuring strictly random choice, sister group-first random choice (the most commonly recommended method), and a forced-broad random choice. Thousands of RAxML runs were completed to gain a large sample size of the results. Both the Noctuoidea and the bees display instability, with the topology found the majority of the time changing based on the number of outgroups. The results obtained when very few outgroup taxa (less than five) were included were found to be the most different from the full results, indicating that studies using few outgroup taxa may be misplacing the root of the phylogeny. Most changes were found within the early arising groups, suggesting that the outgroup taxa may be attaching to different places in the unrooted phylogenies, possibly due to long-branch attraction.