Date:         Tue, 8 Dec 1998 00:13:02 -0600
Reply-To:     "Nichols, David" <nichols@SPSS.COM>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From:         "Nichols, David" <nichols@SPSS.COM>
Subject:      Re: Test for coillinearity in logistic regression

A minor correction: multiple variables with large values on the same _small_
eigenvalue(s) are indicative of the problem sources for collinearity.
 
David Nichols
Principal Support Statistician and
Manager of Statistical Support
SPSS Inc.
 
> -----Original Message-----
> From: Field A [SMTP:A.Field@RHBNC.AC.UK]
> Sent: Monday, December 07, 1998 7:53 AM
> To:   SPSSX-L@UGA.CC.UGA.EDU
> Subject:      Re: Test for coillinearity in logistic regression
>
> Re-run the analysis using linear regression (specifying the same
> dichotomous
> outcome and relevant predictors). To avoid meaningless output switch off
> the
> default options (such as parameter estimates) but click on 'collinearity
> diagnostics' in the statistics dialogue box of the regression options. The
> output should provide you with tolerance values and variance inflation
> factors (VIF). Both Menard (1995) and Myers (1990) concur that a tolerance
> less than 0.1 (or corresponding VIF of greater than 10 - the tolerance is
> simply 1/VIF) is cause for concern. You should also get a table of
> condition
> indexes which lists eigenvalues and the variables in the model. The thing
> to
> look for are variables that both have large values for the same
> eigenvalue.
> If several variables have high values for the same eigenvalue then they
> are
> causing the collinearity problems indicated by the VIF/tolerance values.
> I hope this is coherent enough to be of use!
>
> Good luck,
>
> Andy
>
>
> > ----------
> > From:         Hans Thore Smedbold[SMTP:hans.t.smedbold@MEDISIN.NTNU.NO]
> > Reply To:     Hans Thore Smedbold
> > Sent:         Monday, December 07, 1998 1:31 PM
> > To:   SPSSX-L@UGA.CC.UGA.EDU
> > Subject:      Test for coillinearity in logistic regression
> >
> > This have probably come up before, but I'm new.
> >
> > I would like to perform a test for multicollinearity in my logistic
> > regression model.
> > How can I do that ? (I'm using SPSS v. 8.0).
> >
> > It would be nice if anybody could give me some advice.
> >
> >
> >