Date of Award
Doctor of Education
Educational Leadership, Research and Technology
Dr. Galen Alessi
Dr. James Sanders
Dr. Mary Anne Bunda
This study described and evaluated a new procedure to assess item bias in a minimum-competency test (MCT). This procedure was thought to be capable of estimating the degree of bias contained in a given test item. This is in contrast to traditional item bias detection procedures which focus on the presence or absence of bias in an item. Furthermore, the procedure was thought to be capable of estimating item bias in such a way that the aggregate of item bias (AIB) could be obtained.
The chi-square was found to be the most practical of the item bias procedures and the AIB uses a member of this family. The AIB computes a phi correlation coefficient, squares this to obtain an estimate of the coefficient of determination which estimates the percentage of item variance which is attributable to the demographic characteristic used to separate the subgroups. The coefficient of determination is multiplied by the variance of the item to obtain an estimate of bias in the item and each resulting number is aggregated across test items to obtain an estimate of total item bias in the test.
The AIB was compared to an adaptation of a traditional item bias detection procedure (SSTD) computed on the same data set. The SSTD was a simple directional count of biased items determined by the chi-square procedure. Seven estimates were obtained for both procedures on subgroups which were randomly constructed. An F-test was made on the ratio of the variances of the two procedures. The AIB was found to be the more stable of the two procedures. Subsequent analyses showed that the AIB and SSTD were consistent in their determination of the direction of bias (against Black students) and this was in agreement with population data showing Black students scoring one standard deviation below White students. A t-test of results within the two procedures across the two score points showed that it was allowable to combine these results in the analyses, i.e., scores one and two points below the score which defines a minimally competent student. The AIB was found to be a promising alternative to traditional bias detection procedures.
Rudolph, Laurence E., "An Investigation of a Procedure to Assess Aggregated Item Bias in a Minimum-Competency Test" (1982). Dissertations. 2540.