Precision–recall average precision
Evaluate positive-class ranking with tie-aware noninterpolated average precision.
Your entries stay in this page’s memory and are not submitted to us. Downloads are generated locally. Reloading or leaving the page clears your entries. Do not enter patient information.
Scores can be any finite numeric ranking, with larger scores meaning more likely positive. Outcomes must be 0 or 1 and both classes must be present.
The starting values are hypothetical. Results update as you edit. Use aggregate or synthetic inputs only.
Calculation results
- Average precision
- 0.75555556
- Positive prevalence baseline
- 0.6
- Positives
- 3
- Distinct thresholds
- 5
How to interpret the output
Average precision sums precision at each distinct score threshold weighted by the increase in recall. This is not trapezoidal PR area. Its baseline depends on positive prevalence; values from populations with different prevalence are not directly comparable. Ties enter as a complete threshold group.
Method referenceSource checked 2026-11-06. Read the reproducible calculator validation report.