Multivariate Statistics: Which is appropriate PCA, DFA, MANCOVA?

#1
Hello there,

I am trying to determine which multivariate statistical techniques are most appropriate to address my biological questions and type of data.

1. My first question attempts to determine if a priori predicted morphological traits (>20) are correlated with an ecological trait (in my case various diet, approximately 10 diet types).

The data collected to answer this question are quantitative measurements of morphological traits, (ie. diastema length, jaw angle, body mass, etc.) and qualitative accounts of diet (ie. herbivore, omnivore, etc) for numerous (>500) samples.

I intend to first accounting for size difference amongst samples by performing a linear regression against a size measure (body size), subtracting the residuals, and using these size corrected measurements for further analysis. Would performing a discriminant function analysis determine if certain morphological traits are correlated with diet if they form distinct groups? Would a PCA be better here? Do I need to perform some kind of MANCOVA to determine difference between diet groups? I am lost.

Any help or suggestions would be appreciated!

Thank you!