Reporting Power for Differential Functioning of Items and Tests

• In Event: Explorations in Differential Item Functioning

Abstract

In reviewing the literature related to power being reported in differential item functioning (DIF) analyses, power is similarly defined as the statistically accepted statement of not committing a Type II error. If a test item indeed exhibits DIF, and the DIF technique does not flag it as a DIF item, this is considered a false negative or statistically speaking, committing a Type II error. Power is calculated today in many DIF studies based on the proportion of correct rejections, when the null hypothesis of DIF is false. DFIT’s uniqueness related to its statistical method Item Parameter Replication (IPR) algorithm, will allow power to be calculated beyond a simple statement related to proportions. Empirically observed power may be determined.

Authors

• Keith Darnell Wright, Georgia State University

• T. C. Oshima, Georgia State University