Detecting item bias with the Rasch model.

The purpose of this article is to introduce the concept of item bias, highlighting the differences between the definition of the term as it is used within Rasch measurement and the definition of the term as it is used in the true-score model, non-model based approaches, or multi-item parameter latent trait models. The discussion continues with a description of alternative methods of assessing item bias within the Rasch measurement framework and discusses the power of these methods to detect the presence of item bias. The discussion concludes with several examples drawn from a number of different mathematics tests. This includes a comparison of the Rasch separate calibration t-test and the Mantel-Haenszel approaches.