Date:         Mon, 21 May 2001 14:51:58 -0400
Reply-To:     Peter Flom <peter.flom@NDRI.ORG>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Peter Flom <peter.flom@NDRI.ORG>
Subject:      Re: sig. test for Cronbach's Coefficient Alpha ???
Content-Type: text/plain; charset=US-ASCII

One wonders whether the reviewer is either a) out to get the author or b) out to lunch.  :-)

Since alpha is usually in the range of .7 and up, at least in the uses I've seen for it, it's hard to figure what is meant by a significance test.  Significantly differernt from WHAT? 0 makes no sense (if alpha is diferent from 0, there are BIG problems); but no other value makes sense, unless the point is to compare the reliability of two different scales.....

Do you have any more details?



Peter L. Flom, Ph.D.
Principal Research Associate
National Development and Research Institutes, Inc.
2 World Trade Center
16th floor
New York, NY 10048

(212) 845-4485
(212) 845-4698 (fax)
Peter.Flom@ndri.org

>>> "David L. Cassell" <Cassell.David@EPAMAIL.EPA.GOV> 05/21/01 02:46PM >>>
Gordon Keeler wrote:
> Anyone have a method for generating a significance test for Cronbach's
> Coefficient Alpha ??? A reviewer has asked for this stat, but I can't
> seem to find any reference on how to do it. Does one exist ???

Wow.  I don't think I've ever seen a significance test for this.  In fact,
what I *have* seen is that an acceptable value for alpha varies from
discipline to discipline - so I don't think that a significance test is
likely to be out there.  Perhaps the reviewer is thinking of an
"acceptable"
level.  You may want to throw Nunnally (1978) at the editors - he suggested
that .70 is a good cutoff value for the reliability coefficient.

Is your reviewer just out to get you?  :-)

Reference: J. Nunnally, 1978, Psychometric Theory, McGraw-Hill Publishers

David
--
David Cassell, CSC
Cassell.David@epa.gov
Senior computing specialist
mathematical statistician