I'd like to run a multivariate analysis (Pavoine et al. 2011) that aims to disentagle the effect of environmental gradients on traits, considering its autocorrelation with phylogeny and space. It uses the RLQ approach and works with both discrete and continuous traits. However, I don't know how to incorporate intraspecific variation in discrete traits. For example, I recorded categories of pigmentation on organs of several amphibians, but for some species, these categories vary among individuals. One way would be including individuals as polytomies at the species level, but this would make the test quite hard to interpret.