| Accurate estimates of relationships are critical to research progress. In educational and personnel selection research, such estimates are typically distorted by range restriction. It is therefore important to correct for this distorting effect. However, while indirect range restriction is by far the most prevalent problem in research, available methods of meta-analysis are based on direct range restriction correction procedure. Using such inappropriate correction approach may result in biased estimates of the true relationships between constructs.;Hunter, Schmidt, and Le (2003, manuscript submitted for publication) recently suggested a new model of indirect range restriction. Based on the model, the researchers developed a new method to correct for the problem of range restriction in meta-analysis. The dissertation is designed to provide a comprehensive assessment for the new method.;First, the basic assumption underlying the new model of indirect range restriction was examined. Results show that when the assumption is violated, the true correlations are underestimated. However, the underestimations are generally less serious than those obtained by the existing method based on direct range restriction corrections.;Just like the existing meta-analysis methods, the new method needs information about artifact distributions (i.e., range restriction on the predictor true score) to correct for their attenuating effects. Since such information is generally not available, a procedure was introduced to enable deriving the range restriction distributions of predictor true scores from several well-established generalized artifact distributions in the literature. The resulting distributions can then be used by substantive researchers to correct for the effect of indirect range restriction.;Third, a series of simulations were carried out to test the accuracy of the new method. Results show that the method provided reasonably accurate estimates of the true correlations across different conditions.;Finally, implications of using the new method in research were explored by re-analyzing data of a published meta-analysis examining the validities of employment interviews. Results indicated that different substantive conclusions could have been reached. It is therefore suggested that many meta-analysis studies in the literature may have underestimated the relationships between constructs. |
