|Date: ||Wed, 9 Feb 2005 09:01:24 -0600|
|Reply-To: ||"Reutter, Alex" <firstname.lastname@example.org>|
|Sender: ||"SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>|
|From: ||"Reutter, Alex" <email@example.com>|
|Subject: ||Re: Goodness of Fit in Ordinal Logistic|
|Content-Type: ||text/plain; charset="us-ascii"|
Both of these statistics should be non-significant to indicate a good model fit, so you should be okay.
However, I should note that neither of these tests are very trustworthy when there are a lot of empty cells in your data. If you're running SPSS 11.5 or later, right-click on one of the pivot tables in the Ordinal Regression output and select "Case Studies". The section on "Evaluating the Model" briefly discusses the problem of empty cells and how they relate to the goodness-of-fit statistics.
> -----Original Message-----
> From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU] On Behalf Of
> Constantina Stamou
> Sent: Wednesday, February 09, 2005 5:34 AM
> To: SPSSX-L@LISTSERV.UGA.EDU
> Subject: Goodness of Fit in Ordinal Logistic
> Dear All,
> As I understand it, in ordinal logistic regression
> Model chi-square and Deviance chi-square may be both
> used to assess goodness-of-fit of the model. Also,
> that Model chi-square needs to be significant whereas
> Deviance chi-square must not for a good model fit.
> What happens when Model chi-square is found
> non-significant, and Deviance chi-square is also found
> non-significant? Which of the two statistics is to be
> followed? This is not a hypothetical question, it is
> the result SPSS produced for one of my datasets.
> Thank you in advance.