ROC curve analysis: Criterion and criterion value
In ROC curve analysis, MedCalc does not simply reports threshold or criterion values, but it reports the criterion values with a comparison sign, turning the criterion value into a criterion or indicator of disease.
For example,
- the threshold or criterion value can be 100, but >100 is the criterion for classifying a subject into the positive group; this is the case when larger values indicate disease;
- for other parameters, the criterion value can also be 100, but <100 is the criterion for classifying a subject into the positive group; this is the case when smaller values indicate disease.
To determine the sign or direction (< or >) of the criterion, MedCalc's decision rule is based on the fact that the AUC cannot be less than 0.5. If the AUC is less than 0.5, the decision rule must be reversed (Zweig & Campbell, 1993).
MedCalc proceeds as follows:
- In a first step, MedCalc assumes larger values indicate disease.
- If the Area under the ROC curve (AUC) is lower than 0.5, this assumption is rejected, the decision rule is reversed, and MedCalc takes smaller values to indicate disease.
- In that case the comparison sign of the criterion value x is reversed and >x becomes <x (more precisely, ≤x).
Literature
- Zweig MH, Campbell G (1993) Receiver-operating characteristic (ROC) plots: a fundamental evaluation tool in clinical medicine. Clinical Chemistry 39:561-577.
