A screening model looks for a serious disease, and missing a sick patient is much worse than a false alarm. Which metric matters most?
Options
- Precision
- Recall (sensitivity) (Correct option)
- Accuracy
- Specificity
Answer
Recall (sensitivity)
Why
Recall is the share of real positives the model finds: TP / (TP + FN). It drops when sick patients are missed.
Precision says how many alarms were right, which matters when false alarms are expensive. Specificity measures how well healthy people are cleared, the opposite concern.
In practice you pick a threshold that reaches the recall you need and then check what precision you pay for it.
Remember
Missing positives is costly: maximise recall.