Hi, I am running a binary Logistic Regression on circa 7000 respondents
using both categorical and continuous/likert scaled data as predictors
and Forward LR/Backward LR to find a suitable model. From the results
the Hosmer and Lemeshow test is telling me that the model is not a good
fit, as my significance level is much < 0.05. However my R-sq value
which shows how much the 'badness of fit' improves as a result of the
inclusion of the predictor variables is 31% for Cox and Snell and after
adjustment 46% for Nagelkerke. The classification table shows me that
without any predictor variables in the model 74.4% of the respondents
are classified correctly yet once my predictor variables have been added
83.3% are classified correctly.
I would like to know if I should be taking any notice of the Hosmer and
Lemeshow test? If the predicted classification is significantly
different from the observed should I be using the model? Or can I still
say that even though the model is not a good fit we can still see that
these predictors have a significant effect on the dependent?
Thanks
Jamie Burnett
Research Analyst
MORI
Email: [log in to unmask]
Tel: 0207 3473338
============================
Disclaimer
This e-mail is confidential and intended solely for the use of the
individual to whom it is addressed. Any views or opinions presented are
solely those of the author and do not necessarily represent those of
MORI Limited.
If you are not the intended recipient, be advised that you have
received this e-mail in error and that any use, dissemination,
forwarding, printing, or copying of this e-mail is strictly
prohibited. If you have received this e-mail in error please either
notify the MORI Servicedesk by telephone on 44 (0) 20 7347 3000
or respond to this e-mail with WRONG RECIPIENT in the title line.
============================
_____________________________________________________________________
This e-mail has been scanned for viruses for MORI by MessageLabs. For further information visit http://www.mci.com
|