I am doing predictive analytics in SAS using proc logistic to predict a binary outcome to guide decision making. One measure of accuracy is the misclassification rate, but in my case the desired model is asymmetric in the sense that a false negative is a significantly worse event than a false positive. Any suggestions on how I can tweak Proc Logistic to try to minimize the rate of false negatives rather than just the total rate of misclassifications? Is there some other tool I could be using?
Asked
Active
Viewed 35 times
1
-
Have you tried including Bayesian Priors? – Reeza Jun 29 '17 at 17:49
-
I have not. It sounds interesting in theory, but how would I implement it in practice? – Rob_in_WI Jun 30 '17 at 18:39
-
I think there's an example in the documentation. – Reeza Jun 30 '17 at 19:20