Diagnostics › Diagnostic accuracy

Plot vs criterion values

Criterion analysis traces sensitivity, specificity, PPV, NPV, accuracy and Youden's J across every cutoff, marking the J-optimal threshold.

What is Plot vs criterion values?

The companion to a ROC plot: instead of showing the (FPR, TPR) trace at every cutoff, this shows each metric as its own curve vs the cutoff. Useful for picking a cutoff that meets specific Sens or Spec criteria rather than the global J optimum.

When should I use Plot vs criterion values?

  • Cutoff selection when J isn't the right criterion (e.g. you need Sens ≥ 0.90 and want the highest Spec achieving that).

What data does it need?

Continuous predictor + binary outcome.

What does it report?

Multi-line plot of Sens / Spec / etc. vs cutoff + Youden-optimal threshold.

How do I interpret the result?

Read horizontally: at a chosen Sens (e.g. 0.90) follow the line to find the cutoff that achieves it and read off the resulting Spec.

See also