Youden’s J Statistic

Youden’s J statistic is a metric used to evaluate a binary classifier across possible decision thresholds. It is calculated as sensitivity plus specificity minus one, producing a value from negative one to one, where one represents perfect separation and zero indicates performance no better than chance under the metric. Analysts often select the threshold that maximizes J when false positives and false negatives are considered equally important. That assumption may not fit medicine, fraud, safety, or other settings with unequal costs and changing prevalence. The statistic should therefore complement precision, predictive values, calibration, subgroup results, and domain consequences rather than serving as the only basis for deployment decisions.