As Paul Swank has already pointed out, the Scheffe test can be
applied to pairwise comparisons and combinations of means.
Kirk's Experimental Design text covers this as does Howell's
"Stat Methods for Psychology" (7th ed -  see pages 394-395)
as well as other sources on experimental design and multiple
comparisons.  If you use "a priori" or planned comparisons,
you can also make combinations of means (in Howell 7th ed,
see p371-377 and note his presentation on orthogonal versus
nonorthogonal contrasts).

In general, SPSS provides the option to use the Scheffe test
in its ANOVA procedures (oneway, UNIANOVA, GLM) but
because it has the lowest level of statistical power, it is usually
used in situations where one has to make a large number of
comparisons and Type I errors are much more costly than
Type II errors.

-Mike Palij
New York University
mp26@nyu.edu

----- Original Message -----
From: Eins Bernardo
Sent: Friday, March 05, 2010 9:23 AM
Subject: Re: GLM Univariate

 Marta wrote: >>>Post hoc tests focus on pairwise differences between means, while theoverall F tests whether there is any difference among any pair ofcombination of means. For instance, the F could be significant becausemean1+mean2 combined differ significantly from mean3+mean4 combined. Nopost hoc test would reveal that.   Is the posthoc you are referring to is true also to post hoc in one-way anova? --- On Tue, 3/2/10, Marta García-Granero wrote: From: Marta García-Granero Subject: Re: GLM UnivariateTo: SPSSX-L@LISTSERV.UGA.EDUDate: Tuesday, 2 March, 2010, 5:21 PM Eins Bernardo wrote:> Hi Marta, et al.>> The problem was already solved.  Thank you, Marta, for the solution.>> I would like to raise the issue based on what I have discovered a few> days back.> What is the statistical explaination of the instances wherein> the ANOVA is highly significant but none of the posthoc test is> significant?>Post hoc tests focus on pairwise differences between means, while theoverall F tests whether there is any difference among any pair ofcombination of means. For instance, the F could be significant becausemean1+mean2 combined differ significantly from mean3+mean4 combined. Nopost hoc test would reveal that.HTH,Marta GG=====================To manage your subscription to SPSSX-L, send a message toLISTSERV@LISTSERV.UGA.EDU (not to SPSSX-L), with no body text except thecommand. To leave the list, send the commandSIGNOFF SPSSX-LFor a list of commands to manage subscriptions, send the commandINFO REFCARD

Try the new FASTER Yahoo! Mail.. Experience it today!