From: Sigurd Hermansen on
OK. SAS Proc Logistic reports different measures of predictive accuracy of the model for estimates and sample observations. The c statistic gives us a point estimate of the area under a ROC curve. My SESUG paper on predictive modeling, http://analytics.ncsu.edu/sesug/2008/MPSF-072.pdf , includes a SAS SQL program that calculates the c statistic. I wouldn't stop there. Accuracy of prediction of sample observations does not tell the whole story.
S

-----Original Message-----
From: SAS(r) Discussion [mailto:SAS-L(a)LISTSERV.UGA.EDU] On Behalf Of Rahul
Sent: Saturday, February 20, 2010 2:08 PM
To: SAS-L(a)LISTSERV.UGA.EDU
Subject: Concordance logistic regression SAS

Dear All,

Can someone help me with the following output while building a mode
using logistic regression:

Association of Predicted Probabilities and Observed Responses
Percent Concordant 66.5 Somers' D 0.335
Percent Discordant 32.9 Gamma 0.337
Percent Tied 0.6 Tau-a 0.115
Pairs 497522982 c 0.668

I want to understand the concept, and use of this statistic and how is
this calculated?

Regards.