Date: Wed, 14 Apr 1999 08:48:01 -0700
Reply-To: glaser@pacific-science.com
Sender: "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From: Dale Glaser <glaser@PACIFIC-SCIENCE.COM>
Subject: Re: Reliability analysis
In-Reply-To: <2.2.32.19990414132219.00f7d66c@sph.emory.edu>
Content-Type: text/plain; charset="iso-8859-1"
Please someone correct me if mistaken, but if a determinant approaches zero
then matrix inversion becomes problematic; the classic example is the case
of singularity (variables that are virtual linear combinations of others);
however, I recall asking about this exact issue a couple of years ago, and
if memory serves me well Dave Nichols response was essentially "don't worry"
about this warning. However, you ask an interesting question: in the
context of factor analysis or other correlational methods, a determinant of
zero implies singularity, hence, variables approaching correlation of 0; is
this the same for reliability? Does a determinant approaching zero in
reliability analysis entail highly correlated variables?
Thanks
Dale Glaser, Ph.D.
Senior Scientist, Pacific Science & Engineering
San Diego, CA
-----Original Message-----
From: SPSSX(r) Discussion [mailto:SPSSX-L@UGA.CC.UGA.EDU] On Behalf Of
William Dudley
Sent: Wednesday, April 14, 1999 6:22 AM
To: SPSSX-L@UGA.CC.UGA.EDU
Subject: Reliability analysis
Whne running reliabilty analysis what does the warning * * * Determinant of
matrix is close to zero:
Indicate about the items making up a scale?
Thanks
Bill
***************************************************
* William N. Dudley, PhD *
* Behavioral Science and Health Education *
* Rollins School of Public Health *
* Emory University *
* Atlanta 30322 *
* Phone 404 727 2447 *
* FAX 404 727 1369 *
***************************************************
| 
