Mantel-Haenszel Methodology for Detecting Item Bias: An Introduction and Example Using the Guide to the Assessment of Test Session Behavior

Publisher Website
Google Scholar
Abstract
This paper presents a conceptual overview of the Mantel-Haenszel (MH) technique for evaluating item bias. It then briefly contrasts MH with Rasch methodology and provides an applied example wherein MH is used to examine possible racial/ethnic and gender bias on the Guide to the Assessment of Test Session Behavior. Results from the applied example show that among the 80 possible instances of bias, 10 were statistically significant. This number barely exceeded the chance rate of 9 significant comparisons. Results therefore align with findings from previous studies that show that test observations display few instances of bias. Implications are discussed.