Categorical
ROC analysis (AUC)
Assesses how well a continuous score separates two states, through the ROC curve and the AUC.
Variables
- Score (continuous)Continuous · Ordinal — 1
- True state (binary)Nominal · Ordinal — 1
Options
- Positive state levelSecond level · First level
- Score directionA higher value indicates the positive state · A lower value indicates the positive state
- Confidence-interval methodOn the logit scale · DeLong (untransformed)
What it reports
Reports the area under the curve with DeLong’s standard error and confidence interval; the operating point at Youden’s J, and every coordinate of the curve.
scipy · formula · Open the accuracy report
When does this apply?
To measure how well one continuous or ordinal score separates two states (diseased and healthy, say). Every observation needs one number and the binary true state that number is meant to distinguish. No cut-off has to be chosen: the curve shows every possible threshold at once.