EDPS 36100 Lecture Notes - Lecture 3: Differential Item Functioning
Document Summary
Occurs when people from different groups (commonly gender or ethnicity) with the same latent trait (ability/skill) have a different probability of giving a certain response on a questionnaire or test. Dif analysis provides an indication of unexpected behavior by item on a test. The validity of the criterion should be examined before examining predictive validity. Test bias is important because if test items function differently for different groups, To investigate whether test scores are affected by different sources of variance in the various subpopulations. To determine if these irrelevant sources of variance give an unfair advantage to the subpopulation. An item is judged unbiased if the iccs are the same for both subpopulations. These sub-pops can be divided into a ridiculously large number. Differential item functioning may exist for the correct answer or for the distractors. How ability is divided can be problematic. Tests are imperfect measures, but scores can be examined statistically for bias---outcomes not so much.