Using Multi-Sample Confirmatory Factor Analysis to Test for Construct Bias: An Example Using the K-ABC

Abstract

Recent research designed to investigate construct bias in the Kaufman Assessment Battery for Children served the additional purpose of illustrating several useful methods for detecting construct bias. The purpose of this reanalysis was to illustrate a new, powerful method for testing the similarity of the constructs measured across groups (and, thus, detecting construct bias): multi-sample confirmatory factor analysis (MCFA). Our reanalysis suggested that the K-ABC measures indistinguishable abilities for Black and White children at ages 7 through 8. Those analyses also showed differences in the attributes measured for Black and White children for ages 9 through 12. Follow-up demonstrated that the differences among that latter group were due to trivial aspects of measurement rather than critical differences in the constructs measured by the test across groups. We argue that MCFA provides a more organized, direct, objective, and complete method for detecting construct bias than do other methods.