“An Investigation of Measurement Invariance in the WISC III: Examining a Sample of Referred African American and Caucasian Students”, 2000-08 (; similar):
The American educational system has been frequently charged with discriminatory practices regarding the treatment of minority groups. Specifically, African American students have been thought to achieve intellectual and academically below other ethnic groups. The misconception of underachievement led to and was reinforced by systematic discriminatory practices such as ability grouping, tracking and overrepresentation in educable mentally handicapped special education programs. One controversial issue has been the overrepresentation of African American students in the special education process. The roles that teachers, school personnel and school psychologists play, from the referral through the assessment given, are crucial to the inquiry of why African Americans experience differential educational outcomes in the public school environment.
To further investigate the trend of overrepresentation, we focused on the intellectual measures given and the presence of construct bias. Specifically, the WISC-III was discussed because of it being the most frequently used IQ measure. One emergent technique to assess measurement invariance has been multi-sample confirmatory factor analysis (MCFA). The purpose of this research study was to conduct a multi-sample confirmatory factor analysis of the WISC-III to determine measurement invariance between African American and Caucasian students.
Using MCFA, the WISC-III scores of 545 African American and Caucasian students in the Hillsborough County Public School System were examined to test the presence of measurement invariance. Multi-sample confirmatory factor analysis provided a more direct comparison in the investigation of factor structure equivalence across groups. A 4 step series of analyses was conducted during which all possible parameters (factor loadings, the factor correlation, factor variances, and subtest unique and error variances) were constrained for both groups. From the results obtained there were no statistically-significant differences in the 2 factor model of the [?] the sample of African American and Caucasian students. Within each series of analysis there were no statistically-significant changes in chi square or decline in model fit for either group. Therefore, the proposed 2 factor model as delineated in the WISC-III manual provided a relatively good fit to the sample data.