02-04-2015 11:01 AM
I'm running PROC LOGISTIC and some of my independent variables are dummy variables. Do I need to include these dummy variables in the CLASS statement? I've run it both ways, including them in the CLASS statement and excluding them, and while I get the same Pr > ChiSq both ways, the estimate is different each way. In fact, when including the dummies in the CLASS statement, it appears that the estimate is in the wrong direction. What's the difference here?
02-04-2015 11:13 AM
Are your dummy variables 1/0? If so then you don't need to include it in the CLASS statement.
The opposite sign on the parameter estimates may be due to the difference in the reference level. You can set your reference level explicitly in the CLASS statement.
CLASS dummy_variable (PARAM=REF REF=0);