Date: Thu, 20 Apr 2006 10:32:53 +0800
Reply-To: j.forbes@ecu.edu.au
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: John Forbes <j.forbes@ecu.edu.au>
Organization: Edith Cowan University
Subject: Re: Multinomial Logistic Regression
In-Reply-To: <OF6C1DA6AB.2D0B8D77-ON85257155.004EAB5B-85257155.004F0FA7@uncg.edu>
Content-Type: text/plain; charset="us-ascii"
Thanks Mark,
My research is examining positive psychological functioning, particularly in
terms of its influence on the development of depression. My participants
have provided responses to a range of instruments at two times, and I'm
trying to use my Time 1 data to predict a participant's depression status at
Time 2 (Not Depressed, Borderline Depression, Depressed). I have been using
both scale scores and item scores as covariates (I must admit, though, that
I'm a little hazy on the use of ordinal data as a covariate). I have data
from 571 participants at Time 1, and 409 at Time 2.
Thanks again for your help - and anyone else who wants to chip in! :)
Cheers .................... John
-----Original Message-----
From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU] On Behalf Of
Mark A Davenport MADAVENP
Sent: Wednesday, 19 April 2006 10:24 PM
To: SPSSX-L@LISTSERV.UGA.EDU
Subject: Re: Multinomial Logistic Regression
There are a variety of reasons why your chi-square is high yet may not be
accurate. Can you tell us more about the the models youa re testing? If
you have removed a covariate from the final model or you are using
non-categorical covariates, your chi-square may not be accurately
reflected. Additionally (I know this is a problem with SEM models),
chi-square is notoriously sensative to sample size.Tell us more about what
you are doing and we may be able to help.
Mark
****************************************************************************
****************************************************************************
*******
Mark A. Davenport Ph.D.
Asst. to the Vice Chancellor for Student Affairs
Office of Student Affairs Research and Evaluation
The University of North Carolina at Greensboro
336.334.5582
M_Davenport@uncg.edu
'An approximate answer to the right question is worth a good deal more
than an exact answer to an approximate question.' --a paraphrase of J. W.
Tukey (1962)
John Forbes <j.forbes@ecu.edu.au>
Sent by: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
04/19/2006 04:48 AM
Please respond to
j.forbes@ecu.edu.au
To
SPSSX-L@LISTSERV.UGA.EDU
cc
Subject
Multinomial Logistic Regression
Hi everyone,
I've received some output that I don't quite understand when I run a MLR
in
SPSS 14.0.1.
The model-fitting information yields a chi-square of 301.67, with a sig.
of
.000 - which I believe indicates that the final model does a better job of
describing the data than one using the intercept only. Nagelkerke's pseudo
R-Square is .704. Finally, the classification table indicates that I'm
correct 96.9%, 46.2%, and 68.6% of the time in terms of classifying
participants into my three groups.
All of the above seems to indicate that the model is going quite well.
However, the Goodness of Fit table indicates a significant Pearson
chi-square (p=.000), while the Deviance value is 1.000. I believe that a
significant result indicates a poor model fit - so I'm receiving
conflicting
information from this table. Having .000 'and' 1.000 doesn't seem right to
me, but everything else appears to indicate that the model isn't too
shabby.
Is anyone able to explain this conflict, and perhaps suggest how I might
be
able to resolve it?
Many thanks .................. John
