SOCR logoBPAD

Model Comparison Explorer

Compare logistic, regularized and neighbourhood models on the same split, with uncertainty attached to every difference.

  • model selection
  • cross-validation
  • uncertainty
  • benchmarking

logistic on External test set

MetricValue
AUC0.663
Sensitivity0.500
Specificity0.885
PPV0.400
NPV0.920

ROC curve

00.10.20.30.40.50.60.70.80.9100.20.40.60.811 − SpecificitySensitivity

Calibration

00.10.20.30.40.50.60.70.80.9100.20.40.60.81Mean predicted probabilityObserved proportion

Confusion matrix

Pred +Pred −
True +22
True −323
Ranking is not calibration
Switch to the flexible ensemble and look at its calibration plot: points cluster far from the diagonal even when its AUC looks competitive. Then flip between Internal and External for each model to see which ones degrade gracefully under distribution shift and which ones collapse.