Small Sample Differential Item Functioning Estimation Using Log-Linear Smoothing With Simultaneous Item Bias Testing (SIBTEST) and Cochran’s Z

• In Event: Explorations in Differential Item Functioning

Abstract

Minimum sample sizes of about 200-250 per group are often recommended for differential item functioning (DIF) analysis. However, there are times when sample sizes for one or both groups of interest are smaller than 200 due to practical constraints. Several previous studies have examined the performance of nonparametric DIF methods under small sample conditions. Unfortunately, the definition of small sample differs across studies, and it is still unclear what methods might work at sample sizes smaller than 200 per group. This study attempts to examine the performance of several seemingly promising methods informed by the current literature at a number of small sample and ability distribution combinations. Effect of item parameters on accuracy of DIF detection is also investigated.

Authors

• Pui-Wa Lei, The Pennsylvania State University

• Hongli Li, Georgia State University