Properties of Rasch residual fit statistics

This paper examines the residual-based fit statistics commonly used in Rasch measurement. In particular, the paper analytically examines some of the theoretical properties of the residual-based fit statistics with a view to establishing the inferences that can be made using these fit statistics. More specifically, the relationships between the distributional properties of the fit statistics and sample size are discussed; some research that erroneously concludes that residual-based fit statistics are unstable is reviewed; and finally, it is analytically illustrated that, for dichotomous items, residual-based fit statistics provide a measure of the relative slope of empirical item characteristic curves. With a clear understanding of the theoretical properties of the fit statistics, the use and limitations of these statistics can be placed in the right light.