Why you do not apply data mining technique to solve this problem. Also, the classic multivariate data analyses could be applied (e.g. discriminant analysis ... etc.)

You may consider geometric morphometrics, which involves using standardized images and landmarks to visualize and test for differences in shape between samples.

This is not a simple question to answer because the methods you mention are very different. Neither should be used without a thorough understanding of how it works and (more importantly) of what it assumes about the evolutionary mode(s) of your lineages. The compare.evol.rates function in geomorph compares Brownian rates. If considered simply in terms of statistical performance, it's a sound method. But it assumes that your data fit a one- or multi-rate Brownian model. If those models do not fit, you won't find out by running the function. You will get a comparison of the Brownian rates and a statistical test to determine if the rates differ. The function does not warn you that the rates are incorrectly estimated, as they would be if the mode is not Brownian. So, before using it, you need to determine whether a Brownian model (single or multiple-rate) fits your data better than alternative models that would call for a different rate formula. I have not used AUTEUR because it was (and may still be) limited to univariate data. If you have 10 morphological traits, you will need to think about how to analyze them in light of the correlations among the traits.

Diste una mirada en el paquete RRphylo de Castiglione y Raia? a depender de la pregunta y si vas a usar las informaciones de las tasas (a nivel del tip) para testar algunas hipóteses puede ser útil. Es un metodo relativamente simple, permite datos multivariados y arboles no ultramétricas (por ejemplo si tenes fosiles). abrazos!