If you have two variables, e.g. Hepatic triacylglycerol concentrations (continuous) and metritis present or absent (categorical), and you want to build a ROC curve, it gives you AUC, however, STATA doesn't show that threshold value that is associated or related for that particular response. Does anyone the code or method to see that in STATA?
Thank you!
Hello.
One option might be to convert your continuous variable to categorical e.g., if your continuous values range from 1 to 50 then categories could be 1-10, 11-20, 21-30, 31-40, 41-50.
Then using roctab, detail command in stata will give you an ROC curve with sensitivity/specificity for these cut offs.
Hi Usman,
You can install the cutpt command by Phil Clayton (SSC). This command estimates the optimal cutpoint for a diagnostic test based on sensitivity and specificity: their product (Liu index); their sum (Youden index), and finds the decision point on the ROC curve closest to sensitivity = 1 and specificity = 1. I hope you find this helpful
- Badges
- Science topic