> Do you know what in this case mahalanobis distance is? > > Is it linear DA which is using mahalanobis distance or is it quadratic DA > which is using mahalanobis distance?
Matija, the 'mahalanobis' option produces separating curves that are quadratic. Just like the 'quadratic' option, it's based on estimating a separate covariance matrix for each group. That's what the "stratified" statement is trying to convey.