Date:         Tue, 29 Feb 2000 22:50:07 +0100
Reply-To:     "Christian F.G. Schendera" <schendera@NIKOCITY.DE>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "Christian F.G. Schendera" <schendera@NIKOCITY.DE>
Subject:      Q UNIVARIATE vs NPAR1WAY testing against zero
Comments: To: SASL <SAS-L@UGA.CC.UGA.EDU>
Content-Type: text/plain; charset="iso-8859-1"

I am stuck. A var DIFF is not normal distributed and N < 25. I want to
compute to its difference against zero. If the N for the var were N > 25
then I could use UNIVARIATE to compute Wilcoxon Signed Rank Test (see
below).
However, a stats books says, UNIVARIATE would produce imprecise
results if N < 25, and recommends to use NPAR1WAY to compute exact scores.
Looking
closely at NPAR1WAY I find that it offers Wilcoxon Ranks for small, but
independent groups, that is, vars with CLASS. So, no way out? Does SAS offer
a PROC that covers a test against zero (or tests the difference of related
VART1 and VART2) for non-norm.distr. data and where N = 10?
Cheers,
Chris


        data DATA ;
         DIFF=VART2-VART1 ;
         run ;
/***************************************************************************
****/
/* residuals not normal distributed: Wilcoxon Signed Rank Test against Zero
*/
/***************************************************************************
****/
        proc univariate data = DATA ;
         var DIFF ;
        run ;