OMP011@DE0HRZ1A.BITNET (paul joachim elz) (02/07/90)
dear colleagues, i have the problem to do a u-test ( developed by Mann and Whitney) on unbalanced groups. does anybody have already solved that problem and a working sas program for it ? Or where could I look in ? it would be possible to compute the T - and U-values by PROC RANK ; but for small samples I could not assume a normal distribution for the z-transformed U value. I would have to look it up in a table - uff | On the other hand, I could offer a sas program for point-biserial correlations using the procedure MEANS and TRANSPOSE. Is this of any use for you ? so long, paul joachim elz (omp011@de0hrz1a) station p 4 ^ rheinische landes- und hochschulklinik < * > hufelandstr. 55 v v d 4300 essen 1 federal republic of germany
HAMER@VCUVAX.BITNET (ROBERT M. HAMER) (02/07/90)
paul joachim elz <OMP011@DE0HRZ1A.BITNET> schribt: >i have the problem to do a u-test ( developed by Mann and Whitney) on >unbalanced groups. does anybody have already solved that problem and >a working sas program for it ? Or where could I look in ? A Mann-Whitney U test is a transformation of a Wilcoxon Rank-Sum test. The two tests test equivelent hypotheses and produce the same p-value with the same data. PROC NPAR1WAY does the Wilcoxon Rank-Sum test.