Anchor point selection: An approach for anchoring without anchor items

Arbeitspapier

For detecting differential item functioning (DIF) between two groups of test takers, their item parameters need to be aligned in some way. Typically this is done by means of choosing a small number of so called anchor items. Here we propose an alternative strategy: the selection of an anchor point along the item parameter continuum, where the two groups best overlap. We illustrate how the anchor point is selected by means of maximizing an inequality criterion. It performs equally well or better than established approaches when treated as an anchoring technique, but also provides additional information about the DIF structure through its search path. Another distinct property of this new method is that no individual items are flagged as anchors. This is a major difference to traditional anchoring approaches, where flagging items as anchors implies - but does not guarantee - that they are DIF free, and may lull the user into a false sense of security. Our method can be viewed as a generalization of the search space of traditional anchor selection techniques and can shed new light on the practical usage as well as on the theoretical discussion on anchoring and DIF in general.

Object Details

• Language: Englisch

References and Relationships

• Bibliographic citation: Series: Working Papers in Economics and Statistics ; No. 2018-03

Classification and Topics

• Classification: Wirtschaft ; Multiple or Simultaneous Equation Models; Multiple Variables: General ; Model Evaluation, Validation, and Selection ; Econometric Software
• Subject: item response theory (IRT) ; Rasch model ; differential item functioning (DIF) ; anchor items ; invariant item clusters

Contributors, Places and Time

• Event: Geistige Schöpfung
• (who): Strobl, Carolin ; Kopf, Julia ; Hartmann, Raphael ; Zeileis, Achim
• Event: Veröffentlichung
• (who): University of Innsbruck, Research Platform Empirical and Experimental Economics (eeecon)
• (where): Innsbruck
• (when): 2018

Further information

• Handle: http://hdl.handle.net/10419/184981
• Last update: 10.03.2025, 11:44 AM CET