ROC curves and c statistics

[Kathryn Gardner]

 
Hi,
 
I was recently advised that classification tables in logistic regression are useless, and that I should use the c-statistic for classification instead. I have therefore used ROC curves in SPSS to get the c-statistic for my logistic regression model, but I was wondering whether c-statistics can be computed for regression models with 3 outcomes (SPSS ROC states that the variable must only have 2 outcomes)? If not, is there another technique for validating regression classification results with a 3 category DV?
 
Thanks
Kathryn
_________________________________________________________________
100’s of Music vouchers to be won with MSN Music
https://www.musicmashup.co.uk/index.html