Statistical tests
DeLong test for correlated ROC AUCs
Implemented
statistics.delongDefinition
Compares the ROC AUCs of two models evaluated on the same observations.
Formula
z = (AUC₁ − AUC₂) / √Var(AUC₁ − AUC₂), variance from structural components
Inputs and outputs
- y_true
- scores of model A
- scores of model B
Returns: StatisticalTestResult
Assumptions
No assumptions beyond valid, aligned inputs of the documented types.
Limitations
No metric-specific limitations are documented yet. Interpret the value alongside the task, data, and other metrics.
Python API
import evalsuite as es
es.delong_test(y_true, y_prob_a, y_prob_b)References
- DeLong, E. R., DeLong, D. M., & Clarke-Pearson, D. L. (1988). Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics, 44(3), 837–845.