SELECTING PREDICTORS FOR MAXIMIZING THE CLASSIFICATION EFFICIENCY OF A BATTERY

A simulation approach, relying on synthetic predictor scores and predi cted performance criterion scores for multiple jobs, was used to deter mine the effect that variables in a personnel classification system ha ve on classification efficiency. Classification efficiency was measure d as mean predicted performance determined after optimal assignment to jobs. Population parameters required to generate synthetic scores and to compute evaluation equations for simulations were obtained from th e U.S. Army's Project A (1990) study. The present study provides a com parison of differential assignment theory (DAT) with general aptitude theory and validity generalization. The theoretical value and practica l usefulness of DAT is supported by the finding that both longer test batteries and the use of Horst's differential validity index to select tests increase potential classification efficiency.