What Methods Detect Potentially Biased Items for High-Stakes Testing?
All test items should measure the same content similarly for everyone. Furthermore, only people’s ability should determine whether they get an item right. If something other than ability determines this then that item needs to be examined because it implies the item may be potentially biased.
Westat’s Emily Diaz, Ph.D., and coauthors from Ohio University have published a new article in the International Journal of Assessment Tools in Education based on their study of ways to detect potentially biased items: Detecting Differential Item Functioning: Item Response Theory Methods Versus the Mantel-Haenszel Procedure.
“This study looked at 3 methods under a variety of conditions to determine how often non-biased items were correctly identified as being nonbiased,” notes Dr. Diaz. “Our findings provide insights on the different procedures and error rates.”