480 – The New Face of Statistics Education
Estimation of an Index of Phylogenetic Correlation Using Bootstrap Simulation Technique
Bahman Shafii
University of Idaho
William James Price
University of Idaho
A common objective of bioinformatic analyses is to assess the similarity of species or genotypic variations. Phylogenetic correlation, lambda, is one index of similarity typically used measuring the deviation of an observed phylogeny relative to a dependent Brownian evolutionary model. Values for lambda are estimated through a generalized linear model assuming a variance-covariance structure that has off diagonal elements scaled by lambda. A value of lambda equal to 1.0 is indicative of the Brownian model, while lambda = 0.0 indicates an independent random process. Traditionally, statistical tests for lambda rely on the assumption of a Normal likelihood, and comparison of estimated lambda values between competing phylogenies has not been addressed. The purpose of this paper is to propose an alternative procedure which relies on the re-sampling methodology of the bootstrap. The bootstrap distribution of lambda is estimated which provides a means of computing confidence limits and hypothesis tests without distributional assumptions. The method will be demonstrated using phylogenetic and metabolomic data related to the host specificity of an insect on Brassicaceae plant species.