If your response is binary and you specify the DIST=BINOMIAL option in the MODEL statement, and if you also specify a CLASS statement, then observations that are missing only on the response are not scored by the OUTPUT statement. Predicted values are not computed for these observations.

There are two circumventions for this problem. The first is to use events/trials syntax in the MODEL statement. For this approach, add a variable to the data set that equals one for all observations. Use this new variable as the trials variable in the model. Create a variable that has value 1 for the event of interest and 0 otherwise. Use second new variable as the events variable. Do not include either variable in the CLASS statement. Rerun the model using the event/trials syntax with these new variables.

The second circumvention is to add the STORE statement in your PROC GENMOD step to save the fitted model. Then use the SCORE statement in PROC PLM to score the observations in the input data set.

Operating System and Release Information