ABSTRACT

Scores on a single test, or its item or subtest scores, or scores on a battery of tests, may be used to predict some further performance that is external to the predictor variables. The measure to be predicted might be a supervisor’s rating of performance in a specific form of employment, and this could require a complex combination of aptitudes. Or items, subtests, or tests might be assembled to predict performance in undergraduate or graduate study, and students could be selected for it on the basis of their profile of predictive scores. Differential weight can be given to the scores used, to achieve optimal prediction. Tests, subtests, or items can be eliminated from an initial set when the evidence shows that they are not contributing enough to prediction.