I am analyzing a dataset which consists of measurements for several different metabolites over a number of species. In each cell, the data are expressed as the percentage of the metabolite j in the total volume measured, for species i. I would like to model this percentage as a response to a categorical predictor, which represents the presence or absence of a certain gene, and moreover I would like to allow for the covariance structure in the data introduced by the phylogeny (evolutionary history, shared ancestry) of the species.
I am aware of the various approaches which exist to fit phylogenetic linear models and phylogenetic least squares (e.g.
gls), but I do not think that these can handle my case (although I have tried with
phylolm() approximating to a continuous measure). I thought perhaps that it may be a case for beta regression, but I am unaware of any packages which can use beta regression and handle covariance structure introduced by the phylogeny.
Any help or suggestions greatly appreciated.