Date:         Fri, 17 May 1996 08:08:00 PDT
Reply-To:     Paul Robertson <pfr@ALLP.COM>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         Paul Robertson <pfr@ALLP.COM>
Subject:      Wilcoxon test

Thanks to John Whittington, Robert Ableson, Lawrence Muhlbaier, and
Anupama Narayan for answering my question concerning the WMW
test and tied-ranks.
 
According to Dr. Narayan of SI, the new exact test (added in v6.9 for
mainframes and 6.11 for windows) produces valid p-values even with
tied ranks. Dr Narayan made this statement in a paper delivered at
SUGI-21.
 
Robert Ableson suggested Rupert Miller's book _Beyond ANOVA:
Basics of Applied Statistics_ (Wiley) as a useful reference for
identification of when asymptotic theory, and the older test, may
not be valid.
 
Paul Robertson