Meeting Abstract
Comparative methods allow biologists to address questions pertaining to macroevolutionary processes, making their use in biology commonplace. However, to infer the mechanisms responsible for the diversity of life, a holistic consideration of the relationships between organisms is necessary. Phylogenetic comparative methods take phylogeny to be of fundamental importance in the analysis of multispecies data. The classic method of Phylogenetically Independent Contrasts (PICs) continues to be widely used to meet the statistical assumption of independence of observations. The method assumes that traits evolve by Brownian Motion (BM), so that non-independence in species’ traits is proportional to the length of time species share a common ancestor and independence corresponds to unshared lineages. When the dependence between traits due to shared ancestry (phylogenetic signal) is different than expected under BM, PICs may still be applied if branch lengths can be transformed to serve as measures of covariance. We hypothesized that transforming branch lengths by estimates of phylogenetic signal would be an effective way of standardizing PICs when trait evolution deviates from BM. We simulated traits evolving by different models along random pure birth trees of varying sizes and estimated two parameters of phylogenetic signal: Pagel’s lambda, and the Ornstein-Uhlenbeck strength-of-stabilizing-selection parameter, alpha. We then transformed branch lengths by these signal estimates to compute PICs. Our results indicate that estimating lambda, more so than alpha, is an effective approach to PICs standardization, as branch lengths transformed by this parameter adequately standardized PICs in about 95% of all simulated cases. Likewise, this approach was successful in standardizing PICs for various empirical datasets.
