Examining the measurement quality of tests containing differentially functioning items: Do biased items result in poor measurement?

This study investigated effects of retaining test items manifesting differe ntial item functioning (DIF) on aspects of the measurement quality and vali dity of that test's scores. DIF was evaluated using the Mantel-Haenszel pro cedure, which allows one to detect items that function differently in two g roups of examinees at constant levels of the trait. Multiple composites of DIF- and non-DIF-containing items were created to examine the impact of DIF on the measurement, validity, and predictive relations involving those com posites. Criteria used were the American College Testing composite, the Sch olastic Aptitude Test (SAT) verbal (SATV), quantitative (SATQ), composite ( SATC), and grade point average rank percentile. Results indicate measuremen t quality of tests is not seriously degraded when items manifesting DIF are retained, even when number of items in the compared composites has been co ntrolled. Implications of results are discussed within the framework of mul tiple determinants of item responses.