home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!uvaarpa!darwin.sura.net!spool.mu.edu!olivea!pagesat!spssig.spss.com!nichols
- From: nichols@spss.com (David Nichols)
- Newsgroups: bit.listserv.spssx-l
- Subject: Re: Getting warning # 11302 on SPSS!
- Message-ID: <C1KxH1.81I@spss.com>
- Date: 28 Jan 93 19:43:48 GMT
- References: <1993Jan27.013309.11935@cc.umontreal.ca>
- Sender: news@spss.com (Net News Admin)
- Distribution: na
- Organization: SPSS Inc.
- Lines: 37
-
- In article <1993Jan27.013309.11935@cc.umontreal.ca> irwin@ERE.UMontreal.CA (Irwin David) writes:
- >
- >I am currently doing analysis on data in order to validate a measure of
- >justification of violence toward children (MJVC). At this phase,
- >I am performing factor analysis on the 72 items included in the measure.
- >Six hundred and twelve persons answered the MJVC.
- >
- >I'm using SPSS on a IRIX 4.0 system
- >
- > I am continuously getting the following warning message:
- >
- >Warning # 11302
- > the correlation matrix is ill-conditioned
- >
-
- We get this question a lot, due to a poor criterion choice for the test
- for ill-conditioning, so I thought I'd post my response:
-
- Check your smallest eigenvalue in the initial statistics part of the
- FACTOR output. If that is clearly nonzero, you can ignore the message.
- The message is based on the value of the determinant of the correlation
- matrix, which is the product of the eigenvalues. The eigenvalues must
- sum to the trace of the matrix (in this case the number of variables).
- Thinking about it it is easy to see that this means that the determinant
- is likely to be very small with a large number of variables with even
- fairly well behaved matrices. Indeed, I've actually managed to produce
- this message using data drawn from pseudo-random samples of uncorrelated
- normal populations by using some large number of variables. Thus, the
- choice of criterion for issuing this message is poor. It has been changed
- for version 5.0, so that the message is issued if the smallest eigenvalue
- is very close to zero, rather than the determinant.
-
- --
- David Nichols Senior Statistical Support Specialist SPSS, Inc.
- Phone: (312) 329-3684 Internet: nichols@spss.com Fax: (312) 329-3657
- *******************************************************************************
- Any correlation between my views and those of SPSS is strictly due to chance.
-