[en] The purpose of this talk is to present a novel approach to detect differential item functioning (DIF) among dichotomously scored items. Unlike standard DIF methods that perform an item-by-item analysis, we consider a logistic regression model including item-group interaction (i.e. DIF) effects of all items simultaneously. The method is based on penalized maximum likelihood estimation of a model with a lasso penalty on all possible DIF parameters. Optimal penalty parameter selection is investigated through several known information criteria (such as AIC and BIC) as well as a newly developed weighted alternative. A simulation study was conducted to compare the global performance of the suggested “lasso DIF” method to the logistic regression and Mantel-Haenszel methods, and to evaluate the different optimal penalty parameter selection methods. It is concluded that for small samples the lasso DIF approach globally outperforms the logistic regression method, and also the Mantel-Haenszel method, especially in the presence of item impact, while it yields similar results with larger samples.
Disciplines :
Education & instruction
Author, co-author :
Magis, David ; Université de Liège - ULiège > Département Education et formation > Département Education et formation
Tuerlinckx, Francis
De Boeck, Paul
Language :
English
Title :
Detection of differential item functioning using the lasso approach