Hi all,

I'm looking to do a phylogenetic generalized least squares analysis of a dataset I have on mammals, their average infection status between two habitat types, and various host traits. This dataset is derived from a systematic meta-analysis with quite specific conditions, so the phylogenetic dataset is for only 10 mammal species, where the outcome (infection difference between two habitats) is based on anywhere between 5 to 20 lines of data per species.

I'd like to ask which (if any) host traits predict this difference in infection after controlling for phylogenetic distance, but am concerned that n of 10 really limits any power. As an optimistic point though, the exploratory analysis I've done so far has found very strong effects with very small p values for 3-4 covariates in the PGLS model. So my question is if my sample is simply too small and model overdispersed or if there is any precedent for PGLS with a smaller sample size being deemed acceptable, given the multiple lines of data that went into the outcome for each species (these are included in the PGLS model as weights).

Thanks very much for your time