Hi Collen,
Have you solved the problem? If it is then please send the answer of your question. Your question is very much interested. I have also faced with similar problem.
with regards,
Madan
Colleen Mckay <[log in to unmask]> wrote:
Hi,
I am using Proc Logistic in SAS to work out which independent variables in
may data set predict a binary dependent variable (predicting event=1,
meaning the probability of the event occuring).
The purpose of the model is not for prediction, but to guide which
varaibles are associated in contributing to the event occuring or not, so
therefore the parameter estimates are of prime importance rather than the
predicted probabilities. I am happy with my parameter estimates and they
seem to make sense and are in the expected direction but I am having
difficulty with the fit statistics.
I am using the lackfit option on the model statement (Hosmer & Lemeshow
goodness-of-fit test) which produces a chi-square value of 453.1406 with
df=8 and p<0.0001 indicating that the model is not a good fit. Also I am
using rsquare option on the model statement which produces an R-square of
0.1437 and a Max-rescaled R-square of 0.1976.
My question is that does it matter that the model does not seem to fit
that well? Are my parameter estimates valid to use? If not what modelling
strategies should I use, so that I get a better fit?
Thanks in advance
Colleen Spence
email: [log in to unmask]
--------------
Madan Gopal Kundu
Room no 18, Sukhatme Hostel
Indian Agricultural Statistics Research Institute
Library Avenue, Pusa
New Delhi 12
India
Web: http://www.freewebs.com/madanstata
mobile: 91-9868788406
e-mail: [log in to unmask]
Send instant messages to your online friends http://in.messenger.yahoo.com
|