Receiver Operating Characteristic (ROC) Curve

Definition

A plot of the true positive rate (sensitivity) against the false positive rate (1 minus specificity) across all classification thresholds for a binary classifier. The ROC curve visualises the trade-off between correctly identifying positives and incorrectly classifying negatives. The area under the curve (AUC) summarises overall discriminative performance, with 1.0 representing perfect classification and 0.5 representing chance.

Why It Matters

The ROC curve and its AUC provide a threshold-independent assessment of a classifier's discriminative ability. Unlike accuracy alone, the ROC curve shows performance across all possible decision thresholds, making it invaluable for comparing models and choosing an operating point that balances sensitivity and specificity for a given application.

Example

A diagnostic test for a rare disease achieves an AUC of 0.89. By examining the ROC curve, the clinician identifies that a threshold of 0.32 yields 84% sensitivity and 78% specificity — an acceptable trade-off for a screening tool where missing cases is costly.

Related Terms

• ROC Area Under the Curve (AUC)
• Significance Level
• Sensitivity Testing

Software Notes

• SPSS: Analyze > Classification > ROC Curve; select test variable and state variable
• R: roc() from the pROC package; e.g., roc(response, predictor)
• Stata: roctab for a single classifier; roccomp for comparing multiple classifiers