01-28-2014 05:57 AM
I have a problem that i cant seem to solve.
I want to compare models to find the optimal for separation of two groups.
I have a measurement that is the result of 5 variables. Each variable has 3 to 4 specific values. My total of combinations is therefore 3x4x3x3x3=324 combinations.
At the moment i am using a proc logistic to find the model with the biggest AUC to get the best sens/spec for separation of groups.
I am using a where statement to keep each variable stable
proc logistic data=sasuser.data;
where var1 in (2)and var2 in (8) and var3 in (0.5) and var4 in (0.3) and var5 in (2) ;
model group(event='1') = index_value / outroc=sasuser.data_rocstats3;
I tried to find a loop statement but it does not seem to work. How do i get sas to calculate the ROC curves for all combinations at once? At the moment i am just trying combinations at random to find the optimum but it must be possible to do a smarter solution