Area Under the ROC curves

[deleted_user]

I have a question about the area under the Roc curves.

I plot the roc curves using the logistic procedure and the %rocplot macro but now i need to compare several Roc Curves. Idon't know how to compare the areas under the curves.

If you know the statistical test that compare several AUC and the way to compute it with SAS it will be very helpfull!!

Tank you

[Doc_Duke]

See this paper:

DeLong ER, DeLong DM, Clarke-Pearson DL (1988). Comparing the areas under two or more correlated receiver operating characteristic curves; a nonparametric approach. Biometrics 44:837-845.

and

http://support.sas.com/kb/25/017.html

(obtained from google search
roc compare site:sas.com
).

[deleted_user]

Thank you for your answer but the two dosuments present non parametric comparison of several correlated Roc Curves. This kind of test is appliable for exemple when different diagnostic tests are performed on the same individuals.

In my case the Roc curves i need to compare are not correlated, i have one diagnostic test performed on different populations, that's why it's a little bit more complicated...

[rcm]

I have tried using the %ROC SAS Macro to generate areas under the curve. I am getting areas under the curves that differ from the c-statistic generated by proc logistic. Is this expected? From some of the examples I have seen from SAS papers on ROC curves, the c-stat is equivalent to the area under the curve generated by the ROC macro.

I am new to ROC curve-related analysis so I would appreciate your input.

Thanks.