On the role of phylogenetic analyses for non-heritable traits

Phylogenetically informed methods have become widely accepted in comparative analyses involving heritable species traits.  These techniques correct for spuriously high levels of association between such traits due to shared inheritance.  Recently, techniques have been developed to allow analyses of pairs of traits only one of which shows phylogenetic patterning, by making use of simultaneously diverging “star-shaped” phylogenies.  In this talk, I will show that such approaches are unnecessary, and thus needlessly reduce statistical power.  Using simulated phylogenies, I will show that no inflation of significance occurs in pairwise analyses where one trait shows no phylogenetic pattern, however strong the pattern of inheritance in the other trait.  I will also discuss the more difficult issue of whether phylogenetically informed analyses are needed where traits are intrinsically not heritable, such as analyses of species distribution or abundance.  In such cases, any apparent phylogenetic pattern in the non-heritable trait must reflect the influence of heritable traits, and thus should be part of the “signal” rather than the “noise”.  In such cases, the prudent course may be to perform standard non-phylogenetic analyses, but to examine residuals for phylogenetic pattern, as a means of detecting the influence of heritable factors not included in the analysis.