Many Thanks!! ... View more

Ignoring other issues and given your coding, simply use a WHERE statement: proc logistic data = test; where Age_cat ne 'age2'; class Age_cat; model y=Age_cat + x1; run; ... View more

Hi, Suppose that we are building logit model using few independent variables. Event variable can be dichotomous or multinominal. Sometimes SAS prompts "Error: Complete Separation" or "Error : Quasi Separation". Its mean dependent Variable can be fully explain by any one independent Variable (Complete Separation) or  can be fully explain by combinations of few independent variables ( Quasi Separation). Example 1 Event   Age 1          25 1         45 1         30 0         18 0         24 0          9 0          21 1         40 0         16 1        60 This Shows Complete Separation becaus Event =1 for age > 24 Example 2 Event   Age   Height 0 16 160 0 18 130 0 24 178 0 21 145 0 45 120 1 38 160 1 55 150 1 30 169 1 40 170 1 25 170 This Shows Quasi Separation because Event =1 for age > 24 and Height>149. To remove Compete Separation or Quasi Separation, 1) find out variable those were responsible for Separation ( Check the Collinearity Matrix and Classification Matrix between dependent variable and independent  variables)  then 2) remove those variables or classify them in group so that Complete/Quasi separation no more exist. Hopefully this will help You. Thanks & Regards Ambrish ... View more

You should still add the new variable into your class statement. ... View more