Can I get the standard error of Somers' D or of the C statistic from PROC LOGISTIC?

[JoséQ]

Hi,

I'm not aware of a way to get it from PROC LOGISTIC.

You can get the Somers' D ASE (Asymptotic Standard Error) from PROC FREQ but PROC FREQ and PROC LOGISTIC analyses are not of the same thing, the PROC FREQ analysis is of the raw, observed data, while the PROC LOGISTIC analysis is of the predicted values from the fitted model.

[Rick_SAS]

Question for @JoséQ : From your response, it sounds like the OP could write the predicted value from PROC LOGISTIC to a DATA set and then ask PROC FREQ for the Somers' D ASE on
tables Y*Pred / measures(cl);

Would that provide the correct ASE for the D statistic?

[Ksharp · Accepted solution]

proc logistic data=sashelp.class;
model sex=weight height/rocci;
run;

[Quodly]

Of course, thank you.

[Quodly]

The standard error of D is then twice the standard error of C, I suppose.

[Ksharp]

Sorry. Maybe @Rick_SAS  @StatDave  knows.

[Rick_SAS]

Yes, that's correct. Somers' D is related to the C=AUC by 

C = (D+1)/2

so 

SE(D) = 2*SE(C)