ES-sim-GLM, a Multiple Regression Trait-Dependent Diversification Approach

Identifying the role of quantitative variables on speciation rates is among the main purposes of trait-dependent diversification methods. ES-sim, a recent simulation-based approach that relies on Pearson's correlations, allows testing trait-dependent diversification for single regression models. Here, we modified this approach to include generalized linear models and two independent variables. To examine the effects of multiple traits on speciation we modified ES-sim and integrated generalized linear models instead of Pearson's correlations. We named the new approach as ES-sim-GLM. We further evaluated how this modified method performs in both single and multiple regression modelling. For this, we analyzed the relationship of speciation rates with geographic range size and snout-to-vent length in 216 species from the family Liolaemidae, a South American radiation of Andean lizards. Based on simulations, ES-sim-GLM for single regression models shows high power, low false discovery rates and is robust to incomplete taxon sampling. ES-sim-GLM for multiple regression models shows lower power but also low false-discovery rates. Both remained computationally efficient. Using Liolaemidae data, we found that larger species but with smaller species geographic range sizes were associated with higher speciation rates. To the best of our knowledge, no study as addressed these relationships in this clade. Our results provide new insights on macroevolutionary methods that should be relevant to all organisms and facilitate future studies that aim to understand diversification patterns across the Tree of Life.