Logistic Baseline: Threshold & ROC Explorer
Fit a logistic baseline on kidney-cohort features and trace how one threshold choice moves you along the ROC curve.
- logistic regression
- ROC
- AUC
- threshold
Summary: Out-of-fold (train)
AUC
0.446
n
90.000
Prevalence
0.122
Sensitivity
0.364
ROC curve
Confusion matrix at current threshold
| Pred + | Pred − | |
|---|---|---|
| True + | 4 | 7 |
| True − | 20 | 59 |
Metrics table
| Metric | Value |
|---|---|
| AUC | 0.446 |
| Sensitivity | 0.364 |
| Specificity | 0.747 |
| PPV | 0.167 |
| NPV | 0.894 |
| Youden's J | 0.110 |
A nearly flat ROC curve cannot be rescued by moving the slider
If the curve barely rises above the diagonal, no threshold recovers both sensitivity and specificity at once — the model's ranking, not the operating point, is the limitation. Compare the out-of-fold curve to the external curve to see how far performance can drift once the model leaves the population it was fit on.