Abstract:

Item response theory (IRT) models are widely applied psychometric models. They can predict responses based on characteristics of items and participants. But the validity in all applications of IRT is dependent on the extent to which the selected model accurately reflects the data at hand (goodness of fit). Only when the IRT model fits the data well, can the advantages and functions of IRT emerge (Orlando & Thissen, 2000). Selection of the wrong model would lead to relatively large error in parameter estimation, test equating, the analysis of differential item functioning and so on, which would result in adverse effect (Kang, Cohen & Sung, 2009). Therefore, it is required to evaluate model-data fit before applying IRT (McKinley & Mills, 1985). Common fit statistics in the field of IRT can be expounded and compared from test fit and item fit, which is very important in the field of psychological and educational measurement and also easily neglected in test analysis process. It has not been found yet that there are any similar published articles. Directions of future research for model-data fit could emphasize simulation and empirical study of this issue. And common fit statistics in the field of cognitive diagnosis could be investigated.

Key words: Item Response Theory, Test of Model-Data Fit, Item Fit, Test Fit