multiresponse permutation procedure