Assessing Differential Bundle Functioning Using Meta‐Analysis

Abstract

AbstractDifferential bundle functioning (DBF) has been proposed to quantify the accumulated amount of differential item functioning (DIF) in an item cluster/bundle (Douglas, Roussos, and Stout). The simultaneous item bias test (SIBTEST, Shealy and Stout) has been used to test for DBF (e.g., Walker, Zhang, and Surber). Research on DBF may have the potential to reveal the mechanism underlying DIF. However, an unresolved issue is the lack of an effect size for DBF, making it difficult to assess and compare the amounts of DBF within and between tests. We propose using meta‐analysis techniques to study DBF. By meta‐analyzing DIF indices, we can examine the heterogeneity of DIF across items, using the weighted average of DIF indices in an item bundle as a measure of effect size for DBF. A Monte Carlo simulation study compared the performance of our proposed effect size for DBF and a new test of nonzero average DIF in an item bundle with that of a DBF test using SIBTEST. When the primary and secondary dimensions were moderately correlated, our proposed effect size for DBF had little bias; the test of nonzero average DIF also showed power and Type I error levels comparable to those of the DBF test.