Browsing by Author "Converse, Patrick D."

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Estimating Operational Validity Under Incidental Range Restriction: Some Important but Neglected Issues
    (Practical Assessment, Research & Evaluation, 2017) Brown, Reagan D.; Oswald, Frederick L.; Converse, Patrick D.
    Operational validities are important to personnel selection research because they estimate how well a predictor in practical use correlates with a criterion construct, if the criterion measure were purged of measurement error variance. Because range restriction on a predictor or predictor composite creates incidental range restriction on the criterion, existing methodologies offer limited information and guidance for estimating operational validities. Although these effects of range restriction and criterion unreliability could be corrected with existing equations in a sequential fashion, proper use of sequential correction equations is not always as straightforward as it appears. This research reviews the existing equations for correcting validities, outlines the appropriate method for correcting validity coefficients via sequential equations, and proposes a new equation that performs a combined correction for the effects of incidental range restriction and criterion unreliability.
  • Thumbnail Image
    Item
    Thinking Ahead: Assuming Linear Versus Nonlinear Personality-Criterion Relationships in Personnel Selection
    (Taylor and Francis, 2014) Converse, Patrick D.; Oswald, Frederick L.
    Recent studies suggest that the form of some personality-performance relationships may be curvilinear, meaning that traditional top-down selection is inefficient in capitalizing on underlying personality-performance relationships. This study examines how mean performance is affected by how well the selection method is aligned with the nature of personality-criterion relationships. A simulation manipulated the linearity or nonlinear inflection point of predictor-criterion relationships, and several selection approaches were implemented that varied in level of congruence with these relationships. Results indicate that incongruence can produce notable decrements in mean performance under some conditions. Some evidence also suggests that decrements can be greater when linearity is assumed but relationships are nonlinear (vs. when nonlinearity is assumed but relationships are linear), selection ratios are smaller, and a single predictor is used.