Examining Sources of DIF in Psychological and Educational Assessment Using Multilevel Logistic Regression

In the last three decades, important progress has been made toward more efficient statistical techniques for detecting Differential Item Functioning (DIF). However, the findings are scant when it comes to explaining DIF. Multilevel regression models can expand the knowledge of DIF causes, specifying a DIF parameter that varies randomly over items and testing hypotheses on sources of DIF shared by item bundles. The present study uses multilevel logistic regression to identify the item characteristics that could explain the presence of DIF in short tests or questionnaires, which are usually used in psychological and educational assessment. The usefulness of the approach is tested on measurements of the attitudes toward science of Spanish and English pupils obtained from the OECD Programme for International Student Assessment database.