I've been handed someone else's data (again). It's measurements of six proteins' levels from 20 human brain samples, 5 each of different stages of a disorder. I'd like to run a multivariate regression, but ...
The data is not multivariate normal. None of the individual response variables is univariate normal. And what could I derive from this analysis, anyway, that most biologists could grasp? The PI wants to ask questions about how the disease alters each protein level. When I try to talk about analyzing them as a profile, his eyes sort of glaze over. In any case, how would I present any profile from a multivariate analysis to lab heads that still don't trust anything fancier than an unprotected t test?
