A generalized K statistic for estimating phylogenetic signal from shape and other high-dimensional multivariate data

Area of Science:

• Evolutionary Biology
• Quantitative Genetics
• Bioinformatics

Background:

• Phylogenetic signal measures trait similarity due to common ancestry.
• Existing methods are limited for high-dimensional multivariate traits like shape.
• Statistical properties of these methods are not well understood.

Purpose of the Study:

• To introduce and validate a generalized K statistic (K(mult)) for high-dimensional multivariate data.
• To assess the statistical performance and reliability of K(mult) using simulations.
• To compare K(mult) with existing methods for evaluating phylogenetic signal.

Main Methods:

• Generalization of the K statistic (Blomberg et al.) to multivariate data (K(mult)).
• Utilizing the equivalency between covariance and distance matrices.
• Computer simulations based on Brownian motion with varying trait dimensions and variation.
• Evaluation of Type I error rates and statistical power for hypothesis testing.

Main Results:

• K(mult) maintains an expected value of 1.0 across varying trait variation and dimensions.
• Squared-change parsimony methods show confounding changes with increased dimensions/variation.
• K(mult) demonstrates appropriate Type I error and high power for detecting phylogenetic signal.
• Method performance is robust across different phylogeny types, species numbers, and covariance structures.

Conclusions:

• K(mult) is a reliable and statistically sound method for quantifying phylogenetic signal in high-dimensional multivariate traits.
• It offers a significant improvement over existing methods, enabling more accurate evolutionary interpretations.
• The method was successfully applied to analyze head shape evolution in Plethodon salamanders.