A Bias Guideline Preempts Experimental DIF Detection

The present article reports on a judgmental analysis of the items in the English subtest of Iranian University Entrance Exam (2009) investigating possible biased items. Plenty of statistical techniques, such as Logistic Regression, Mantel-Haenszel method, and IRT approaches, are developed to detect differential item functioning (DIF) and biased items. They require a pilot study of the test with a sizeable number of subjects. However, sometimes pre-testing is not possible and only subjective and judgmental analysis can be used to detect potential biased items. Off course, judgments should be informed by research findings and experts’ opinions. This study suggests that research findings and experts’ opinion can be combined to create a bias guideline for language test development. If research shed enough light on the issue of biased items and tests, a bias guideline may preempt the experimental DIF detection. This study utilized previous research findings, experts’ opinion on bias, and the author’s intuition to propose a bias guideline and figure out the possible biased items or bundles (groups of items) in the Iranian University Entrance Exam.