The problem that there are more morphological variables than samples in some groups (four individuals in tour case) is probably common. I usually complete a preliminary PCA with the hope that the scores on the first few axes show good (or retain some) discriminant power. I would like to hear more opinions from the list, and the possible caveats in using such a procedure. This circumvent the problem that some variables are linear combination of other (alternatively, what about the taboo stepwise variable selection?)
On the other problem (ontogeny and sex confounding the effects of species), the problem is that possibly there are significant interactions between sex and species, and ontogeny and species (the latter could imply to violate the assuptions for a common PCA approach). I am also very interested in knowing more details or opinions on how to manage this issue. Miquel Palmer == Replies will be sent to list. For more information see http://life.bio.sunysb.edu/morph/morphmet.html.
