APPLICATION OF THE MULTIPLE DISCRIMINANT FUNCTION TO DATA FROM THE AIRMAN CLASSIFICATION BATTERY

Abstract
A new approach to the selection and classification of AF trainees is described. The new procedure, which is directed primarily toward the establishment of differences in abilities required for various jobs, utilizes the multiple discriminant function. The study involved 6105 airmen who had satisfactorily completed training in 1 of 8 specialties and had also completed all the tests of the Airmen Classification Battery AC-1. Multiple discriminant analysis of this data revealed that essentially all the information concerning the separation of the 8-specialty centroids in the 17-dimensional Airman Classification Battery space is described by 2 linear combinations of the 17 variates. These 2 functions were believed to represent mechanical ability and intellectual ability. A study of the distributions of 2 discriminant scores, computed for each airman, showed that the assumption of bivariate normality was tenable for the discriminant score distributions with but one exception. Centour scores were computed for each group under the assumption of a normal bivariate discriminant score distribution. These scores would be helpful for career guidance of new airmen except for the large overlap of the 8 groups in the Airman Classification Battery space.