Selection for Multiple Jobs from a Common Applicant Pool

摘要

Procedures for selecting recruits from a common applicant pool to makeassignments to a set of 9 or 14 MOS (as surrogates of job families) are evaluated in an unbiased simulation design. Synthetic test scores are generated based on Project A data. Five levels of an assignment strategy level ranging in complexity from one in which jobs and individuals are considered in random order to one which approaches an LP algorithm in both complexity and efficiency. A sixth level is also considered, a primal LP algorithm, as a baseline against which to compare mean predicted performance (MPP) scores provided by the other multiple job assignment procedures. Least squares estimates (LSEs) of the criterion, separately for all 6 strategy facet levels, use 28 Project A tests as predictors. LSEs are used as assignment variables when the best' weights are obtained from a back sample and as evaluation variables from which to compute MPP when the weights are obtained from the designated population. Two types (levels) of minimum cut scores, one closely resembling the Army operational cut scores with regard to range and height, and the other proportional to dual parameters, are used in conjunction with the 6 levels of the strategy facet. Two sets of assignment variables (AVs), with and without the effect of Brogden's removed, are compared. AVs based on LSEs are also compared with AVs derived from three different types of a single factor. A consistent increase in MPP is found as the complexity of the multiple job selection algorithms approaches the complexity of the LP algorithm.