Date:         Fri, 25 May 2007 08:50:45 -0400
Reply-To:     Don Asay <donald.asay82@gmail.com>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Don Asay <donald.asay82@gmail.com>
Subject:      Comparing Predictive Ability for multiple Logistic Regressions
Content-Type: text/plain; charset=ISO-8859-1; format=flowed

I am running multiple logistic regressions using the same independent
variables, but changing the dependent variable. I was wondering if there is
a way to test if the model is a better predictor for some dependent
variables than others. I am looking at the log liklihood and the cox and
snell R-squared, and the model I created works well for some dependent
variables Cox & Snell =.483 and not so well for other dependent variables
Cox & Snell = .124

I was wondering if there was a statistical technique or and SPSS option that
would allow me to say that the model predicts much better for this dependent
variable than the other dependent variable. Visibly it appears that one is a
much better predictor, but this could be due to a large number of other
things, so I would like to use some sort of statistical test. I have about 5
dependent variables I am using this model to test, most clump together up in
the .3 to .5 range but one is an outlier and I am wondering if it is
different from all the others.

Any help or reccomendations on the topic would be greatly appreciated.

Don