Improving Classification Efficiency by Restructuring Army Job Families.

摘要

This report presents one of a series of studies designed to improve classification efficiency by applying principles of differential assignment theory. The major objectives of this study are to examine classification efficiency as a function of the number of job families and to examine alternative methods for clustering or forming job families. Factors investigated include alternative methods for constructing assignment variables or predictor composites; the effects of using a more economic criterion, size, and heterogeneity of the test battery from which assignment variables are formed; and the size of analysis samples used to form assignment variables. Sets of synthetic scores (entities) that have the statistical properties of empirical test scores were generated. The synthetic entities were assigned under the differing experimental conditions being simulated, and mean predicted performance was computed for each alternative being investigated to form the unit of comparison among alternatives. We refer to the simulation of personnel systems process using synthetic scores as model sampling. A cross validation design was employed that eliminated traditional back sample inflation due to sampling error. In one design, 18 jobs from Project A, validated against core technical proficiency criteria, were used in a model sampling experiment. In a second design, 60 jobs were used validated against skill qualification tests.