Date: Thu, 21 May 1998 12:29:56 -0500
Reply-To: "Nichols, David" <nichols@SPSS.COM>
Sender: "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From: "Nichols, David" <nichols@SPSS.COM>
Subject: Re: Unadjusted means and adjusted means are reversed in MANCOVA
You may want to check exactly what's being reported in terms of the
transformed variables and the contrasts applied to the between factor. The
default contrast is DEVIATION, which would produce the first group minus
the mean of the two groups at the between level, with the transformed
variable for the within contrast being the normalized version of a
difference score of second time point minus first time point. This could be
the issue: if you use DEVIATION(1) or SIMPLE(1) contrasts (using the first
category as the left out or reference category, respectively in the
contrasts), you'll get an estimate that goes in a different direction.
David Nichols
Principal Support Statistician and
Manager of Statistical Support
SPSS Inc.
nichols@spss.com
----------
From: Larry W Hawk [SMTP:lhawk@ACSU.BUFFALO.EDU]
Sent: Monday, May 18, 1998 12:25 PM
To: SPSSX-L@UGA.CC.UGA.EDU
Subject: Unadjusted means and adjusted means are reversed in MANCOVA
A colleague just asked about a problem with a data analysis. I'm not sure
on all the details, but here's the crux of it.
2 group x 2 time design, with a third level of time as the covariate. (Not
sure if multiple dependent variables are in the model)
Problem is that the covariate-adjusted means for the crossed interaction
are in the opposite direction of the unadjusted means.
I suggested that she check whether the ANCOVA assumptions were met, namely
the assumption that the relationship between the DV and the covariate was
consistent (i.e, does not interact with) any other factor in the design.
My suggestion seemed less than satisfactory (to me). Other ideas...
Thanks,
Larry Hawk
--------------------------------------------------------------------
/ Larry Hawk, Ph.D. / /
/ Assistant Professor / Phone: 716-645-3650 x231 /
/ Department of Psychology / Fax: 716-645-3801 /
/ Park Hall / E-mail: LHAWK@ACSU.BUFFALO.EDU /
/ SUNY-Buffalo / /
/ Buffalo, NY 14260-4110 / /
--------------------------------------------------------------------
