Differential Item Functioning Analysis of a Statewide Visual Arts Assessment Using a Two-Stage Procedure

• In Event: 66.077 - Methodological Considerations of Rasch Models

Abstract

One of the key components in educational tests is to detect biased or Differential Item Functioning (DIF) items for the evaluation of fairness and validity. A two-stage procedure utilizing Rasch Trees and Mantel-Haenszel (MH) approaches for characterizing DIF based on different combinations of subgroups was performed in this study. A total of 1,463 fourth graders’ multiple choice items assessed by a large-scale visual arts assessment in the 2017-2018 academic year were analyzed. The Rasch Trees results show an interaction DIF between gender and ethnicity for the global model. In addition, two “C” items and three to 14 “B” items were identified by the MH method, suggesting that some items may demonstrate slight to moderate or moderate to large differences.

Authors

• Kelvin T. Pompey, University of South Carolina

• Ning Jiang, University of South Carolina - Columbia

• Yin Burgess, University of South Carolina

• Ashlee A. Lewis, University of South Carolina

• Jingtong Dou