A Comment on Phylogenetic Correction

Abstract

Statistical methods are now commonly used to take into account the expected lack of independence of observations across different species (due to their phylogenetic relatedness) when computing correlations or regressions among traits. The methods are often interpreted as removing that part of the regression or correlation that is an artifact due to phylogeny and there is an expectation that the corrected regression or correlation coefficients will usually be closer to zero. It is shown here that this is not an accurate way to interpret these methods. The effect of taking phylogeny into account is to reduce the variance of the estimated regression or correlation coefficients. Their means are not because since estimates of regression coefficients are unbiased whether or not the correct phylogeny is taken into account. Estimates of correlations are only slightly biased (and in the opposite direction that many expect).

Skeletal correlates of quadrupedalism and climbing in the anthropoid forelimb: Implications for inferring locomotion in Miocene catarrhines

“…Thus, it is important to estimate the degree of phylogenetic signal in the correlation of skeletal traits and locomotor behaviors to ensure that the relationship is investigated under the proper regression model. If the chosen model is incorrect, analyses can lead to higher Type I error rates and invalid results (Freckleton et al, 2002;Rohlf, 2006).…”

“…In order to determine the model of best-fit for each behavior and prediction, we tested four different models including ordinary least-squares (OLS), phylogenetic generalized least-squares Phylogenetic generalized least-squares regression employs a variance-covariance matrix in which the diagonals represent the estimated branch lengths from the root of the tree to each species, and the off-diagonals represent the branch length shared by pairs of species from the root to their last common ancestor (Rohlf, 2006;Spoor et al, 2007;Walker et al, 2008). This variance-covariance matrix is used in a weighted least-squares regression that incorporates covariance related to phylogeny (Rohlf, 2006;Walker et al, 2008).…”

“…This variance-covariance matrix is used in a weighted least-squares regression that incorporates covariance related to phylogeny (Rohlf, 2006;Walker et al, 2008). All pGLS regressions are run in REGRESSIONv2.M in Matlab (available from T. Garland, University of California Riverside, Riverside, CA).…”

“…When building predictive models of this kind based on a range of taxa, it is essential to test whether one should incorporate inferred phylogenetic relationships and estimated branch lengths in order to take into account covariation between closely related species. Thus, the present study employs phylogenetic generalized least-squares (pGLS) to investigate potential phylogenetic structuring of the data (Rohlf, 2006;Walker et al, 2008).…”

The Expensive Brain: A framework for explaining evolutionary changes in brain size

“…Ricklefs and Starck 1996;Martin et al 2005), especially if there are many closely related species of similar body size in a genus, and is therefore no guarantee to obtain the most reliable results for a particular dataset (Rohlf 2006). Thus, we always report results of analyses based on log-transformed species data (TIP, denoted here as "raw") as well as from independent contrasts analyses (PIC, denoted here as "IC").…”