Date:         Sat, 23 Mar 1996 09:58:38 -0500
Reply-To:     "William B. Ware" <wbware@EMAIL.UNC.EDU>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From:         "William B. Ware" <wbware@EMAIL.UNC.EDU>
Subject:      Re: Proper Post hoc test using Kruskal-Wallis Test
Comments: To: Linda Serra Hagedorn <hagedorn@UIC.EDU>
In-Reply-To:  <199603221532.KAA72310@mailserv0.isis.unc.edu>

On Fri, 22 Mar 1996, Linda Serra Hagedorn wrote:
 
> I have run a Kruskal Wallis one way analysis of variance using SPSS for the
> mainframe.  The result of this test was a significant difference among the
> four groups.  To determine where significant differences lie, I must perform
> some sort of post-hoc analysis.  My question is what and how?  Since the
> Kruskal-Wallis is a non-parametric test can I just perform a series of
> pairwise Mann-Whitney comparisons?  Or, would that threaten the overall
> error rate?
 
Zwick (In Keren and Lewis (1993). A Handbook for Data Analysis in the
Behavioral Sciences:  Statistical Issues.  Lawrence Erlbaum) shows a
procedure for post-hoc MCPs following a Kruskall-Wallis test.
 
 
z(sub i)= (Rbar(i) - Rbar(j))/sqrt{(n(n+1)/12)(1/n(i) + 1/n(j)}
 
The critical value is given as sqrt{Chi-squared} for df=k-1 at the upper
tailed alpha experimentwise...
 
 
______________________________________________________________________________
 
William B. Ware, Professor and Chair                   Educational Psychology
CB# 3500                                               EMAIL:   wbware@unc.edu
University of North Carolina                           PHONE:   (919)-966-5266
Chapel Hill, NC      27599-3500                        FAX:     (919)-962-1533
______________________________________________________________________________