Dummy Interaction Term in Proc Logistic

[Yan]

I got different results when running logistic regression using Proc Logistic with interaction term.

One way, I directly put the cross product term in the model statement of Proc Logistic; the other way I created an interaction variable in DATA step. Both variables in the interaction term are categorical variables.

Results are different when I included the missing in the analysis. One is significant, the other is not. Can't figure out why this will be different. Not sure how SAS deal with the interaction term in Proc Logistic.

[Doc_Duke]

You should not get different results. I do it routinely and get the same either way. I suspect your DATA step code is making a subtle assumption difference.

[Yan]

Thanks for your reply.

I did get the same result if I excluded the missing from the analysis. But different when I include the missing values in the analysis.

[plf515]

Please post your code so that we can attempt to diagnose the problem