Fairness in Computerized Testing: Detecting Item Bias using CATSIB with Impact Present
Man-Wai Chu, Hollis Lai · Alberta Journal of Educational Research · 2014
In educational assessment, there is an increasing demand for tailoring assessments to individual examinees through computer adaptive tests (CAT). As such, it is particularly important toinvestigate the fairness of these adaptive testing processes, which require theinvestigation of differential item function (DIF) to yield information about itembias. The performance of CATSIB, a revision of SIBTEST to accommodate CATresponses, in detecting DIF in a multi-stage adaptive testing (MST) environmentis investigated in the present study. Specifically, the power and type I error rates on directional DIF detection of an MST environment when positive and negative impact, group ability differences, was investigated using simulation procedures. The results revealed that CATSIB performed relatively well in identifying the items with DIF when characteristics of the group and items were known. Testing companies are able to use these results to enhance test items which provide students with fair and equitable adaptive testing environments.