Module 19: Differential Item Functioning
This module is intended to prepare the reader to use statistical procedures to detect differentially functioning test items. To provide background, differential item functioning (DIF) is distinguished from item and test bias, and the importance of DIF screening within the overall test development process is discussed. The Mantel-Haenszel statistic, logistic regression, SIBTES'r, the Standardization procedure, and various IRT-based approaches are presented. For each of these procedures, the theoretical framework is presented, the relative strengths and weaknesses of the method are highlighted, and guidance is provided for interpretation of the resulting statistical indices. Numerous technical decisions are required in order for the practitioner to appropriately implement these procedures. These decisions are discussed in some detail, as are the policy decisions necessary to implement an operational DIF detection program. The module also includes an annotated bibliography and a self-test.
Keywords: differential item functioning, item response theory, logistic regression, Mantel-Haenszel, SIBTEST, policy, standardization procedure