Date:         Wed, 30 Oct 2002 15:23:34 EST
Reply-To:     Rcarlstedt@aol.com
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Rcarlstedt@aol.com
Subject:      Correction for Attenuation
Content-Type: text/plain; charset="US-ASCII"

Can one justify correcting for attenuation on the basis of the reliability of
an outcome measure, that is, a criterion variable? For example, I would like
to argue that correlations between select predictor variables (in this case
personality and behavioral measures) are/could be greater than observed
because said criterion measures did not necessarily reflect potent
(meaningful) outcome measures (e.g., a global measure of performance).

I know correction for attenuation can be applied when the reliability of a
predictor variable (i.e., reliability of a test instrument) is less than 1.0
(always the case of course). However, how would one justify (if at all) the
application of this correction method based on the relaibility of a
criterion/outcome measure. Otherwise is there an alternative when focusing on
the reliability of an outcome/criterion measure?

Thanks in advance.

Roland A. Carlstedt, Ph.D.



______________________________________________________
Roland A. Carlstedt, Ph.D.
Chairman, American Board of Sport Psychology*
Assistant Psychologist and Staff Clinical Psychotherapist,
Behavioral Associates of New York City,
www.americanboardofsportpsychology.org
+1 917-680-3994

* An American Psychological Association Approved Provider of Continuing
Education for Psychologists