Comparative Study of Some Item Bias Detection Methods / Abdullah A. AI-Qataee
Several methods of item bias detection indices were introduced in the last decades. The IRT methods were said to be more elegant compared to others. However, due to some practical difficulties, the search for an alternative method is necessary. The purpose of the present study was to compare the ability of several methods to detect item bias compared to the two parameter IRT models. Thus, DIF methods, the transformed item difficulty (TID) method, chi-square methods, and item discrimination indices were compared using data from the information Subtest for the Normative Sample of the Saudi version of the WISC-R. Sex biases were considered. The results indicate that IRT methods are more elegant than other methods, thus confirming previous findings. Chi-square methods were found to be a good substitute for IRT methods. More specifically, Cammilli chi-square was most effective followed by MH, and Angoff modification of the TID. The new modification of the item discrimination (ID) index seemed to be more effective than the original one. However, further study of the new modification of ID is needed.