The Spectrum of Invariance: Measurement, Prediction, and Selection Invariance in Certification Testing

• In Event: Rating and Scoring Issues in Assessment

Abstract

A key component in the process of validating psychological and educational measures is to ensure that the measure being used has scores that are invariant across demographic groups. In testing, invariance can has a spectrum of meaning spanning measurement, prediction, and selection bias. Borsboom et al. (2008) demonstrated that, under certain conditions, measurement invariance and selection invariance may be incompatible, leading to a conundrum for psychometricians. This research investigates the full spectrum of invariance in a selection/certification context and the compatibility of selection invariance and measurement invariance in an applied setting. Results suggest that measures may essentially demonstrate measurement, prediction, and selection invariance in applied settings.

Authors

• Andrew Jones, American Board of Surgery

• Jonathan D. Rubright, American Institute of Certified Public Accountants