An iterative technique for clustering criteria which retains optimum predictive efficiency

This paper describes an iterative technique, programmed for an electronic computer, that at each step r duces the number of criterion clusters and provides optimal weights for the tests and a cost table can be used to determine the appropriate balance between predictive efficiency and number of criteria clusters.

Abstract

Abstract : IN A PERSONNEL CLASSIFICATION PROGRAM, MAXIMUM PREDICTIVE EFFICIENCY RESULTS FROM APPLYING LEAST-SQUARES WEIGHTS TO A SET OF PR DICTOR TESTS TO PRODUCE A SEPAR T COMPOSITE SCORE FOR EACH CRITERION. With a large, complex organization such as the Air Force, it is not feasible to compute composite scores to predict the success of every individual in every criterion situation. This paper describes an iterative technique, programmed for an electronic computer, that at each step r duces the number of criterion clusters and provides optimal weights for the tests. A cost table can be used to determine the appropriate balance between predictive efficiency and number of criterion clusters. Solution of a sample problem illustrates each st p of the technique. (Author)