Model Selection and Goodness of Fit for Phylogenetic Comparative Methods

• Activity Number: 170
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Dwueng-Chwuan Jhwueng*+
• Companies: National Institute for Mathematical and Biological Synthesis
• Keywords: phylogenetic comparative methods ; Independent Contrasts ; phylogenetic mixed model ; Ornstein-Uhlenbeck (OU) process ; spatial autoregressive model ; Akaike information criterion

Abstract

Phylogenetic comparative methods (PCMs) have been applied widely in analyzing data from related species. Many such methods have been proposed but their fit to data is rarely assessed. We assess the fit of several phylogenetic comparative methods to a large collection of real data sets gathered from the literature. We also compare the models using model selection criteria. Results show that Felsenstein's Independent Contrast and the independent, non-phylogenetic, models provide better fit for most real data. We then fit those methods to bivariate data sets and compare correlation estimates and confidence intervals from each model using simulations. We find that correlations from different models are often qualitatively similar so that actual correlations from real data seem to be robust to the PCM chosen for the analysis. Therefore, while there is no evidence that the parameter-richer models fit better, using them does not overly change the result in most bivariate analyses.