chapter  4
16 Pages

GAUSSIAN DISTRIBUTIONS OF SIGNAL AND NOISE WITH UNEQUAL VARIANCES

Unfortunately for the experimenter, not all detection tasks involve signal and noise distributions with equal variances. In these cases we cannot infer the entire ROC curve from a single pair of hit and false alarm probabilities but need a set of points for the curve. This means that while a single yes no task is sufficient to estimate in the equal variance case, a series of yes-no tasks, or a rating scale task, is necessary to determine sensitivity when variances are not equal.