Effect of Sample Size, Number of Biased Items, and Magnitude of Bias on a Two-Stage Item Bias Estimation Method

A two-stage procedure for estimating item bias examined with six indexes of item bias and the Mantel-Haenszel (MH) statistic; the sample , the number of biased items, and the magnitude the bias were varied. The second stage of the did not identify substantial numbers of positives (unbiased items identified as biased). , the identification of true positives in the stage was useful only when the magnitude the bias was not small and the number of items was large (20% or 40% of the test). weighted indexes tended to identify more true false positives than their unweighted item theory counterparts. Finally, the MH identified fewer false positives, but did not small bias as well as the item response indexes. Index terms: differential item functioning, bias, Mantel-Haenszel statistic, two-stage estimation.