Considerations on model inference
Phylogenetic multiple regressions allowed us to evaluate the unique
effect of each predictor on FST while correcting for
phylogenetic autocorrelation, which had not been accomplished in
previous broad-scale studies. Additionally, we note that after adding
the factor latitudinal region, the scaling parameter that corrects for
phylogenetic autocorrelation (λ fit in Table 1) became insignificant.
This suggests that latitudinal region decreases the phylogenetic
autocorrelation in the residuals modeled by our phylogenetic regressions
(Freckleton, 2009). In fact, an alternative across-species multiple
regression of model 7 (i.e., a linear model assuming phylogenetic
independence) yielded identical results with indistinguishable fit to
the data (ΔAIC=1.9). We suspect that region captured important
phylogenetic information in FST and species traits;
within each regional species pool, lineages share strong biogeographic
and phylogenetic affinities. Put another way, we think that regional
affiliation is the most important underlying factor influencing
FST values at a global scale, and when not included,
phylogenetic signal becomes a proxy for latitudinal region due to the
tendency for closely related species to occur in similar regions.