Title: Anchor Methods for DIF Detection in the Rasch Model

Authors: Julia Kopf, Achim Zeileis, Carolin Strobl

Abstract: Differential item functioning (DIF) indicates the violation of the
measurement invariance assumption, e.g., in psychological tests. For item-wise
DIF detection using item response  theory (IRT), a common metric for the item
parameters of the groups that are to be compared (e.g. for the reference and
the focal group) is necessary. In the Rasch model, therefore, the same linear
restriction is imposed in both groups. However, the question how the items in
the restriction - termed anchor items - are selected appropriately is still a
major challenge.

For clarity, we first propose a conceptual framework for categorizing anchor
methods: The anchor class describes characteristics of the anchor methods and
the anchor selection strategy guides how the anchor items are located. Second,
we propose a new anchor class termed the iterative forward anchor class that
iteratively includes items in the anchor. Third, we investigate new anchor
selection strategies and combine them with two different anchor classes. Our
results show that the new suggestions allow for a lower false alarm rate and a
higher hit rate compared to previously suggested methods. Hence, items with
and without DIF are classified more accurately and the evaluation of the
invariance assumption in the Rasch model can be improved. An implementation in
R is currently under construction.