ABSTRACT
In differential item functioning (DIF) studies, examinees are classified into a reference group (r), or a focal group (f), or are ignored as members of other groups (o), based on their responses to a group ide_B.tification question. Some examinees fail to answer this question and so introduce missing data that can affect the inferences that are based on DIF measures. Because DIF often is concerned with a comparison of an item’s performance for a relatively small subpopulation (the focal group) versus its performance for a much larger subgroup (the reference group), even a small amount of missing data may have profound effects. For example, self-ide_B.tified Black examinees commonly make up about 6% of the population taking the insurance practitioner licensure test used in our example, and nonresponse to the racial/ethnicity question is about 9%. Thus, if a major part of the nonresponding population is Black, the introduction of this large number of nonrespondents into the Black group can have a substantial effect on any inference made.
