Date:         Mon, 10 Jun 2002 17:32:39 -0400
Reply-To:     "Zdaniuk, Bozena" <BZdaniuk@UCSUR.PITT.EDU>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         "Zdaniuk, Bozena" <BZdaniuk@UCSUR.PITT.EDU>
Subject:      covariates and mediation
Content-Type: text/plain; charset="iso-8859-1"

hello, I have a statistical question about mediation. I am familiar with the
Baron and Kenny mediation model using regression. My question is: can
covariates effects in a RM ANCOVA be inerpreted as indicating (lack of)
mediation? Details:
I have a grouping IV A and a within-subject IV B. I find A x B interaction
effect on DV. There is a reason to believe that variables C, D, E might
mediate the impact of IV's on DV. So I add three covariates (C, D, E) to the
model and the interaction effect is still there with the eta sq. pretty much
unchanged. Can I conclude that C, D, E do not mediate the relationship
between IV's and DV?
Thanks in advance for any help you might offer on this issue.
Bozena Zdaniuk
University of Pittsburgh