Package pROC does the job - roc.test with parameter paired=F
Karl
?
----- Urspr?ngliche Mail ----
Von: Karl Knoblick <karlknoblich at yahoo.de>
An: r-help at stat.math.ethz.ch
Gesendet: Montag, den 27. Juni 2011, 12:19:48 Uhr
Betreff: Differences between ROC curves
Hallo!
Can anybody help with the comparison of ROC curves? I want to compare two
uncorrelated?ROC curves, i.e. the data sets of the two ROC curves are from
different samples.
One way to compare ROC curves seems to be the comparison of the AUCs. Although I
found the package pROC with he comparison of correlated ROC curves, I did not
find anything about the comparison of uncorrelated ROC curves.
Does anybody know how to compare the AUC of wo uncorrelated ROC curves?
(Why do?I want to do this? I have two samples an older and a newer one.The
question is, if the diagnostic test (AUC of ROC curve) is better in either the
older or newer sample or if they are equal.)
Thanks for any help!
Karl