Phylogenetic relationships between populations of pupfish
A caveat for the presented methods and results is that we did not correct for phylogenetic relationships when estimating or comparing P matrices across populations and species. Comparative methods are important tools for making comparisons between samples that violate assumptions of independence within comparison groups, such as comparing closely related species that vary in their evolutionary relatedness to each other. While the pupfish species and populations used in this study are indeed closely related, we would argue that the patterns and interpretations presented here are still informative for several reasons. SSI generalists from both radiating and generalist-only ponds have generally been shown to be more closely related to one another than to specialist pupfish species– especially the scale-eaters– or to the Caribbean pupfish populations used in this study (Richards et al. 2021, Martin and Feinstein 2014, Martin 2016), a pattern that is reflected in appendix figure A1. This information suggests that we would expect a high level of similarity in the P matrices of SSI generalist-only and SSI radiating groups due to their phylogenetic relationship, yet our methods detected significant differences between these two groups. A similar argument can be made to justify the interpretation that the P matrix of the SSI generalist-only group is different from the P matrix of the Caribbean group. Finally, the comparisons between the Caribbean and SSI radiating groups indicates relative similarity between the two P matrices. Additionally, the expectations of similarity in P matrices due to evolutionary history are the same for generalist-only and radiating groups. The fact that we detected differences when comparing SSI generalist-only and Caribbean matrices suggests that the observed similarity between the SSI radiation and Caribbean groups is not a false positive.