People often make decisions about a person or object’s group membership based on a fallible indicator measure. For example, a teacher who sees a bruise on a child may need to decide whether to report this as a possible case of child abuse. Suppose the school has a check sheet for rating injuries that yields a larger score for injuries that are more likely to indicate child abuse. However, there is overlap in distributions of scores for cases that are due to accidents and cases that are child abuse, so there is risk of a wrong classification. It is important to identify cases of abuse; failure to report the case could be very costly if indeed the child is a victim of abuse. On the other hand, reporting the case as abuse when the bruise came from a sports injury could also be costly to the child and the child’s family.
In some classification situations, costs and benefits can be calculated in financial units (e.g., dollars), but often benefits and costs are subjective, and are referred to generally as ‘utility.’
UTIL is an Excel program that calculates cutting scores on an indicator measure to maximize classification utility. UTIL takes into account the population base rates, means and variances of the indicator variable, the positive utility of correct decisions, and the negative utility of incorrect decisions.
Citation: Berger, Dale E. (2018). Maximizing Utility in Group Classification with UTIL2.1. Retrieved from https://wise.cgu.edu