A proposal for categorizing the severity of non uniform differential item functioning: The polytomous case

Abstract

Differential item functioning (DIF) is present when something about the characteristics of a test taker interferes with the relationship between ability and item response. Non uniform DIF (NUDIF) exists when there is interaction between ability level and group membership. The aim of this study is to propose guidelines to evaluate the severity of NUDIF, taking into account the family of Rasch models. The severity of NUDIF is evaluated trying to quantify the overall biasing impact of the presence in the test of NUDIF items, on the estimated measure of the latent trait of interest.