Date:         Mon, 3 Nov 2003 19:02:50 -0500
Reply-To:     arnoldr@uwindsor.ca
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Robert Arnold <arnoldr@uwindsor.ca>
Subject:      posthoc tests in analysis of covariance
Content-Type: text/plain; charset=US-ASCII

Hi. I'm not used to working with ANCOVA, so there may be an easy way to do
what I would like to do, but haven't seen how to do.  I would like to
compare
mean scores on a family functioning measure for people in  different
locations, controlling for their education and family income.
Specifically,
I would like to compare those in sites 1 and 2 with those in site 5 and
those
in site 3 with those site in 4.

The comparison is easy without the covariates.  I can simply enter

  ONEWAY famfunc by site with edlevel income
    / posthoc = <name of a chosen test> .

The output compares each site with each other, and I can examine the
comparisons of interest, while ignoring the rest.

According to the documentation, the posthoc subcommand is available
for UNIANOVA (the GLM univariate analysis command).  When I try to
use it, though, I get a message saying that the POSTHOC command is
being ignored because there are covariates in the model.   UNIANOVA
will print out the differences between each site and a chosen reference
site, but not output from POSTHOC.

Is there some other way to get comparisons between specific groups
after controlling for covariates?

Thanks in advance for your help.