Publication at Faculty of Mathematics and Physics, Faculty of Education |
2019
Abstract
Many methods for detection of differential item functioning (DIF) are derived from comparison of item characteristic curves (ICC) including approaches based on IRT models and non-IRT methods such as logistic regression or its extensions. However, both approaches assume parametric model for probability of correct answer.
In this talk we introduce general nonparametric approach of comparison of regression curves proposed by Srihera and Stute (2010). We further adapt the method and we apply it to estimate ICCs and to test for DIF.
For that purpose, we propose several weight functions which can improve DIF detection process and we compare their properties via simulation study.