Ramblings about phylogenetic eigenvector regression

After lots of reading and mucking around in r, the best (and possibly only) way I can get phylogenetically independent contrasts and maintain species as data points, is by variance partitioning using Phylogenetic Eigenvector Regression (PVR). I had some email contact with Thibaut Jombart from Imperial College in London and he recommended this regression approach. Its reasonably robust as it deals with any type of trait data (binary, categorical etc)and does not assume any explicit evolutionary model. See this paper for a good example of what’s possible: http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0089314. Theoretically you could partition ecological effects from the data (i.e. guild), but I’m worried that the resultant residual would be devoid of biological meaning (but it could be interesting….). I could also keep going down this road and look at niche conservatism and the true phylogenetic component alone…. it is a rabbit hole though and I want to keep to the original question (what is the effect of leaf litter and shade on community phylogenetic and functional structure).

 What I’m currently thinking is to take the species PVR residuals, multiply them by log(abundance) to generate plot scores, and then do factorial ANOVAs to test for litter/shade effects in my data set.  As an ecologist trying to incorporate phylogeny,  there is still no clear methodological pathway to follow and each method has pitfalls. 

 

Advertisements