Phylogenetic Analysis of Covariance by Computer Simulation

Biologists often compare average phenotypes of groups of species defined cladistically or on behavioral, ecological, or physiological criteria (e.g., carnivores vs. herbivores, social vs. nonsocial species, endotherms vs. ectotherms). Hypothesis testing typically is accomplished via analysis of variance (ANOVA) or covariance (ANCOVA; often with body size as a covariate). Because of the hierarchical nature of phylogenetic descent, however, species may not represent statistically independent data points, degrees of freedom may be inflated, and significance levels derived from conventional tests cannot be trusted. As one solution to this degrees of freedom problem, we propose using empirically scaled computer simulation models of continuous traits evolving along «known» phylogenetic trees to obtain null distributions of F statistics for ANCOVA of comparative data sets