07-08-2015 11:03 PM
What could go wrong here? Thank you so much. Truly appreciate the response.
07-09-2015 08:54 AM
Could you post your code? (And data too, if possible.)
07-09-2015 09:26 AM
Nothing goes wrong but SAS does not output OR's when interaction terms are included in the model.
If the interaction term is significant you may (must) present stratum specific OR's and you can get those by the ESTIMATE statement.
(If the interaction term is not significant, you remove it from the model again and will get the OR's you need).
PROC LOGISTIC DATA=one DESCENDING;
CLASS e f / PARAM=REF REF=FIRST;
MODEL d = e f e*f;
ESTIMATE 'OR for E in stratum F=1' e 1 e*f 1/EXP;
ESTIMATE 'OR for E in stratum F=0' e 1 e*f 0/EXP; * or just e 1;
ESTIMATE 'OR for F in stratum E=1' f 1 e*f 1/EXP;
ESTIMATE 'OR for F in stratum E=0' f 1 e*f 0/EXP; * or just f 1;
Hope this helps.
07-09-2015 10:56 AM
To elaborate further, the reason that this is happening is that without an interaction term, the odds ratio is constant for a variable. When adding an interaction, the odds ratio changes depending on the values of the interacted variables. So you need to tell SAS the values of the interacted variables where you want an odds ratio estimate.
07-09-2015 11:29 AM
It is possible that attempting to post your entire question in the subject will make it impossible to respond. There seems to be a bug related to the length or content of the titles.
Better would be a subject similar to:
Missing odds ration with interaction terms in proc logistic
and them provide additional details about what you were trying to accomplish , what code you tried and what did not work as expected.