Date:         Fri, 25 Mar 2005 14:37:25 -0500
Reply-To:     Jeff Stuewig <jstuewig@gmu.edu>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Jeff Stuewig <jstuewig@gmu.edu>
Subject:      computing mediated effect and sobel
In-Reply-To:  <42445FA9.7060508@jura.uni-hamburg.de>
Content-type: text/plain; charset=us-ascii; format=flowed

         When computing the mediated effect (a*b or c - c` ) can one use
either the unstandardized b weight or the standardized beta?  Additionally,
can you use either for computing the sobel test?  I was always taught to
use the unstandardized b weight and its corresponding standard error but in
Howell's book "Statistical Methods for Psychology" he uses the standardized
beta and its standard error.  Are these equivalent?  Thanks.


         Jeff