Ok, so I’ve got this problem. In some work, I’ve got a sample in which entities, let’s call them subjects, are rated positive or negative. The sample of subjects is then filtered by two criteria, resulting in a 3d plot: “aplha cutoff”, “beta cutoff”, and “positive rate”. This is all well and good, but how can I make an ROC (receiver operating characteristic) curve from this? Would it have to be either in 4d or using a set of 3d surfaces? Is there an expressable way to do it? I just can’t think…
Maybe I can get something nice by doing quadrature on a scatter of false pos vs. false neg… does that make sense?
Thanks for all the help guys! 
I’m going with the scatter plot. It looks great, but the damn EPS file is huge.