Interpreting the relationship between item difficulty and DIF: Examples from educational testing

Abstract

The investigation of differential item functioning (DIF) is commonplace in educational testing. Most often it results in no more than the discarding of a proportion of the developed items. Apart from the potential threat to validity of the instrument due to differential loss of certain types of items, this also is inefficient in terms of item development costs. This paper discusses certain situations where particular patterns of non-uniform DIF can be related to item difficulty, and modelled by enhanced Rasch models. This allows for a more efficient usage of items, as well as, potentially, an improved understanding of the complexities of how educational tests function across different sub-po..