Tutorial: Maximizing Utility in Group Classification

Michael R. Healy1, Dale E. Berger1, Christopher L. Aberson2, Amanda Saw1, & Victoria L. Romero1

1 Claremont Graduate University.  2 Humboldt State University


Practitioners often make predictive decisions about a person’s group membership based upon test performance. For example, a clinical psychologist might need to determine if there is sufficient evidence to diagnose a patient as being depressed or a cognitive gerontologist might need to determine if an older adult is suffering from natural age-related memory impairment or a more serious condition such as Alzheimer’s disease. These classification judgments can be difficult to make given natural variability within a population, costs of correct and incorrect decisions, and base rates for a person’s true group membership.

The WISE Util applet calculates cutting scores to maximize classification utility. The program takes into account population base rates and means and variances of the test score distributions, and the benefit of a correct classification and the cost of an incorrect classification. The program interactively displays the model’s underlying weighted normal distributions, cutting scores which designate predicted group membership, and expected classification accuracy.

The classification applet appears below. A further description of the applet is available on the next page of this tutorial.

Begin the tutorial

1,857 total views, 1 views today