Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | Richard Goldstein <richgold@ix.netcom.com> |
To | statalist@hsphsun2.harvard.edu |
Subject | Re: st: Comparison of logistic regressions with different outcomes |
Date | Tue, 04 Jun 2013 16:54:42 -0400 |
it is not completely clear what you mean by "better at predicting outcome A or outcome B"; however, one place to start is with some of the post-estimation procedures; e.g., the area under the ROC curve (-lroc-) which compares the models in terms of which does a "better" job of discrimination; sounds like you should start with the manual entry on logistic postestimation if this is the right direction for you, you might also want to look up the suite of commands for -roc- analysis; start with -h roc- if this is not the right direction (maybe you're more interested in calibration?), you need to give us a little more guidance on what you want Rich On 6/4/13 4:44 PM, Zachary Neal wrote: > I am looking for a way to compare two logistic regression models > estimated on the same data, with the same predictors, but with > different outcomes. In essence, I want to know whether a given set of > predictors is better at predicting outcome A or outcome B. > > I found this site useful for understanding the pseudo-R2 statistics > reported by the fitstat command: > http://www.ats.ucla.edu/stat/mult_pkg/faq/general/Psuedo_RSquareds.htm > As I understand it, for all their faults, the same pseudo-R2 can > validly be compared across two models estimated on the same data, with > the same outcome, but with different predictors. But, my case is the > opposite of this: the predictors remain the same across models, but > the outcomes change. > > Any suggestions would be appreciated. > * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/faqs/resources/statalist-faq/ * http://www.ats.ucla.edu/stat/stata/