Evaluation of the Quality of Nine Item-Fit Statistics of Rasch Model and Statistics Criteria Used in the Northwest Evaluation Association Item Calibration Procedure

• In Event: Evaulating and Exploring the Nuances of Item and Person Fit

Abstract

The general purpose of any fit statistic is to provide information on an item’s goodness-of-fit (GOF) to a given model. This study provides empirical evidence on the quality of NWEA fit statistics and fit statistics criteria used in NWEA item calibration procedure and underscores the importance of understanding the effects of manipulated factors (IRT model, sample size, person distribution, and misfit rate) on classification accuracy of fit statistics under given cutoff values. Among all fit statistics NWEA used in its calibration procedure, four of them, INFIT, OUTFIT, Rpbim and PexpPM are the most promising fit statistics because, in general, they have the best classification accuracy on flagging unfit items.

Authors

• Shudong Wang, NWEA

• Gregg Harris, Northwest Evaluation Association