A Tool to Optimize the Predictive Accuracy of Personnel Selection and Classification Instruments

The primary goal of this proposal is to design a method for providing unbiased estimates of test validities, intercorrelations, and means for Navy applicants when the test validation samples contain only selected individuals. In practice, personnel selection test and criterion performance scores, obtained only from Navy selectees, rather than the entire population of applicants are used to estimate the validity of the Armed Services Vocational Aptitude Battery (ASVAB) tests and composites. Since the individuals in these samples have been preselected in a variety of ways and are restricted in their range of ability, their predictive validities do not accurately estimate applicant validities. Accurate estimates of applicant test validities are required, since these tests are actually designed to make selection and classification decisions for applicants. Although there are known and scientifically acceptable procedures for estimating applicant validities, involving correcting validities observed on the selected samples, they do not take into account the effect of sequential selection systems that can result in samples which fail to meet the required statistical assumptions of the available correction procedures.During the Phase I effort, we will investigate the feasibility of designing a methodology to estimate, for Navy applicants, or any population, an unbiased predictor and criterion intercorrelation matrix. In short, our major goal is to provide unbiased and accurate estimates of school validities in the applicant population, with its biases and distortions, rather than in a general population of non-applicants, such as the ASVAB norming group, the American Youth Population (U.S. Department of Defense, 1983).