MODEL ADEQUACY TESTS FOR AN EARLY HIGH DISPARITY PHYLOGENETIC COMPARATIVE MODEL

Brownian Motion (BM), Ornstein-Uhlenbeck (OU), and Early Burst (EB) are considered canonical models in phenotypic phylogenetic comparative methods. Here, I expand the repertoire of evolutionary modes by combining EB and OU models to produce an ‘Early High Disparity’ (EHD) model. Given a time-calibrated phylogeny and simulated trait data, the EHD model was tested with implementation in the R packages PCMBase and PCMFit. Model adequacy (MA) analyses for EHD are presented to assess how well the generating (simulating) models matched the best-fitting models. Two main model adequacy tests were performed for the relative fit of the EHD model: 1.) fitting the EHD model to BM, OU, and EB simulations of trait data; 2.) fitting BM, OU, EB, and EHD models to EHD-simulated data. Degree of ‘best-fit’ was assessed using ΔAIC_c for all model adequacy tests. For the first two model adequacy tests, all instances of the incorrect model retrieved as best-fitting were tallied, with the expectation that the incorrect model is retrieved as best-fitting 15% out of all simulated datasets tested. The EHD model often performed poorly when fit to BM, OU, or EB generated simulations (1-15% type 1 error rate for EHD to EB model fits; 0-6% type 1 error rates for EHD to OU model fits). For the second set of MA analyses, with a total 9 combinations of α and ρ for the EHD simulations, error rates ranged from 6-9%. These results indicate relatively good performance of the EHD model to discern generating modes of evolution from fitted modes. At least with simulated datasets, the EHD function used in this work may accurately capture signals of early high disparity. Testing with empirical datasets is a crucial next step, including tests with marine reptiles such as ichthyosaurs, which seem to display patterns consistent with an EHD evolutionary mode.