Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | Ashwin Ananthakrishnan <ashwinna@yahoo.com> |
To | statalist@hsphsun2.harvard.edu |
Subject | st: calculating AUROC when odds ratio cannot be defined |
Date | Mon, 16 May 2011 08:24:37 -0700 (PDT) |
Hi, I'm trying to calculate the area under the ROC for a model where the outcome is binary (disease 1 vs. disease 2) and the predictors are response rates at various time points (also binary variables - for example, response at 2 months - Y/N, response at 3 months - Y/N). For the 3 month cut off point, all patients with disease 1 respond (100%) while about 40% of those with disease 2 respond. Obviously, I can't calculate an odds ratio for this time point. Is it possible to calculate an AUROC with disease 1 vs. disease 2 as the outcome of interest and response at 3 months as the predictor variable? Or because there are no events in one of the 2x2 boxes, is it that an AUROC cannot be calculated? Thanks, Ashwin * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/