topic Comparison of C statistics between 2 Cox proportional hazards models in Statistical Procedures

Comparison of C statistics between 2 Cox proportional hazards models

Nagi — Tue, 23 Feb 2021 10:13:04 GMT

Hello everyone,

I am trying to compare c statistics between 2 models based on Cox proportional hazards model. I found out the way to calculate each c statistics of Cox models like below.

proc phreg data=nm concordance ;
model death_year*death(0) = sbp_m1;
run;

proc phreg data=nm concordance ;
model death_year*death(0) = sbp_m2;
run;

I want to compare 2 models: Y(death) = a + b･X(sbp_m1) and Y(death) = a' + b'･X'(sbp_m2) using c statistics.

I think that there should be a statistics test which tests the null hypothesis; "c statistics (Model 1) - c statistics (Model 2) = 0". But I have no idea about the corresponding code.

Does anyone know it?

Any comments are welcome.

Sincerely,

Nagi

Re: Comparison of C statistics between 2 Cox proportional hazards models

OsoGris — Tue, 23 Feb 2021 13:44:27 GMT

I don't believe there is any code that can compare two c-statistics from two different models here.

Re: Comparison of C statistics between 2 Cox proportional hazards models

Nagi — Tue, 23 Feb 2021 23:14:47 GMT

Thank you for your reply!

So, is there no way to test a difference between c statistics?

I have found p value between them in some papers. How did they calculate it?

Many thanks,

Nagi