Signal Detection Theory

d' separates sensitivity from response bias — building the ROC curve

Parameters

Detection Metrics

d' = 1.50
Hit rate P(Y|S)
False alarm P(Y|N)
d'
Beta β (bias)
A' (AUC approx)
d' = z(Hit) − z(FA). Pure sensitivity measure — unaffected by criterion shift. ROC curve sweeps over all criteria.