Some observations on multivariate statistical methods in anthropological research