Monday, July 26, 2010

Research Briefs 7-26-10: CHC theory & measures (WJ III) invariant (no psychometric bias) across blacks and whites

Kane, H. D., & Oakland, T. D. (2010). Group Differences in Cognitive Ability: A CHC Theory Framework. Mankind Quarterly, 50(4), 318-331.

Abstract

The Cattell-Horn-Carroll (CHC) model of cognitive ability as represented in the Woodcock-Johnson Tests of Cognitive Ability-III (WJ-III) was examined for Black and White adults matched on various demographic variables. Although Whites performed higher than Blacks (i.e., race differences were found in test scores and accompanying factor means), the results of multisample confirmatory factor analyses found that the same constructs are measured in different groups. Therefore results are directly comparable, and in this sense measured differences can be interpreted as “real” differences on the dimensions that the test is meant to measure.
Part of authors conclusions:
With respect to the primary purpose of this study, although White-Black differences in cognitive ability are affirmed in favor of Whites, these analyses reveal no source of psychometric bias (i.e., differences in loadings, test intercepts, and error variance). Constructs are represented adequately and without undo influence of error. The structural fidelity of the WJ-III factor model is psychometrically sound, making it a suitable instrument for psychologists when estimating general and broad cognitive abilities for individuals and groups. The reported indices of fit (e.g., TLI, GFI, and RMSEA) suggest that the threestratum CHC model fits the WJ-III data fairly well and provides evidence of construct validity. This finding substantiates a growing body of research literature that upholds the WJ-III as a strong representation of CHC theory (e.g., Edwards & Oakland, 2006; McGrew & Woodcock, 2001). Further, the data support Carroll’s (1993) belief that the CHC theory is essentially invariant across racial-ethnic groups. Notably, the group differences in test performance are smaller than in most other studies (e.g., Osborne & McGurk, 1982). This particular result is likely due to the samples being matched by parental education and occupational status. In the US, this kind of control is expected to remove approximately one third of the Black-White difference that may be expected in demographically representative samples (Jensen, 1998;
Lynn, 1998).

Technorati Tags: , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , ,