Estimating unrestricted population parameters from restricted sample data in employment testing

Abstract

This study examined the accuracy of Alexander, Alliger, and Hanges’ (1984) method for estimating unrestricted univariate predictor means and variances from sample data drawn from three populations in two personnel selection contexts: (1) where there was direct nonstrict truncation on the predictor, and (2) where there was direct strict truncation on the predictor. In addition, the accuracy of corrected (estimated unrestricted) validity coefficients based on estimated population predictor standard deviations was assessed in the nonstrict truncation condition. In general, there was inconsistency in the accuracy of the population predictor mean and standard deviation estimates obtained across the present datasets and conditions. Caution is advised in the interpretation and reporting of corrected validity coefficients in employment testing based on estimated population predictor standard deviations. Index terms: employment testing, personnel selection, range restriction, true validity estimation, unrestricted population parameters.