An Item Bias Investigation of a Standardized Aptitude Test.
MINNESOTA UNIV MINNEAPOLIS DEPT OF PSYCHOLOGY
Pagination or Media Count:
Verbal and quantitative data from a standardized aptitude test SCAT, Series II, Level 2 were analyzed separately for Native American and White high school students. Item correlation matrices were factor analyzed for each group, separately for each ability. Coefficients of congruence comparing factor structures between groups were high for the first verbal factor and the first and second quantitative factors, implying that ability factor structures were similar for the two groups. The first factors were of sufficient size to allow parameterization of the items by item characteristic curve ICC methods. Item difficulty b parameters derived for the two groups were compared by regressing difficulty parameters for the Native American group on the difficulty parameters for the White group, and values of elliptic-D were computed for each item and group. Results led to the conclusion that there were no reliably biased items in the verbal subtest, while there were two reliably biased items in the quantitative subtest--one item biased against the Native American group and one biased against the White group Internal consistency reliabilities were higher for the Native American group in both tests, and the scores of Native American students were better predictors of high school rank than were scores for the White students but these results were significant p.05 only for the quantitative subtest. Results indicated that different approaches to the identification of bias led to different conclusions. Thus, additional research is needed to determine which indices of item and test bias yield the most meaningful approach to the analysis of basis in ability tests. Author