AUC comparison between different models

[Nagi]

Hello, I am a beginner of SAS and I am struggling with AUC comparison between 2 different model.

In SAS, I can compare AUC between the nested(hierarchical) models using ROCCONTRAST. But, how about the different models like the following two logistic regression models?

1 . where sex = 0;

     model cvd = bmi;

2.  where sex = 1;

     model cvd = bmi;   

The have the same independent variable but under different condition.

Is it possible to compare and test the difference in that case?

Thanks in advance,

Nagi

[Ksharp]

Check the last curve .

data F;
 set sashelp.class(where=(sex='F'));
 y=ifn(_n_>4,1,0);
run;
data M;
 set sashelp.class(where=(sex='M'));
 y=ifn(_n_>4,1,0);
run;


data test;
 set sashelp.class;
 y=ifn(_n_<4,1,0);
 if _n_<10;
run;
data test_f;
 set f test(drop=y);
run;
data test_m;
 set m test(drop=y);
run;



proc logistic  data=test_f;
model y=age weight height;
output out=pred_f(where=(y is missing)) p=pred_f;
run;
proc logistic  data=test_m;
model y=age weight height;
output out=pred_m(where=(y is missing)) p=pred_m;
run;


data want;
 merge test(keep=y) 
 pred_f(keep=pred_f)  pred_m(keep=pred_m);
run;
proc logistic data=want;
model y=pred_f pred_m/nofit;
roc 'F' pred=pred_f ;
roc 'M' pred=pred_m;
run;

[Nagi]

Thank you for your reply.

In IFN function, what does "_n_>4" mean?

Thanks in advance,

Nagi

[Ksharp]

I just made some dummy Y . i.e. the first 4 obs have y=1 ,others have y=0 .

[Nagi · Accepted solution]

Thank you for your reply!

I could solve this problem. I appreciate your help.

Nagi