Research Seminar
Detection of differential item functioning: Crossed random-effects models and robust diagnostics toolsDavid Magis and Paul De BoeckKU Leuven | |
| Abstract: | When assigning a test to a set of examinees, it is very important to detect whether differential item functioning (DIF) occurs; an item is said to show DIF if its difficulty level is not identical among all (subsets of) examinees taking the test. After a brief review of some well-known techniques (Mantel-Haenszel test, likelihood-ratio test, standardized p-DIF method) we turn to a model-based approach as a useful method to detect DIF. Two models are considered: the random item profile (RIP) model, with different difficulty levels for all subsets of examinees; and the random item mixture (RIM) model, where item difficulties are defined by a mixture of item parameters and a set of latent DIF and non-DIF item classes. In this talk we shall mostly focus on the RIP model, for which DIF detection rules have not yet been developed. We present two such rules, based upon robust diagnostic investigation tools, and we highlight their usefulness by means of a simulation study. Eventually, all DIF detection methods are illustrated on the basis of a real data set. |
| Date: | Tue Sep 30, 12:15 pm - 1:15 pm |
| Place: | room 00.60 (Department of Psychology, Tiensestraat 102, 3000 Leuven) |