CAJ | 학술논문

本文将IRT常用资料一模型拟合检验统计量x2和G2引入认知诊断领域，具体讨论了这两个统计量在认知诊断资料一模型拟合检验的可行性及其侦查效果，为研究者及实际应用者在认知诊断资料模型拟合检验中提供借鉴及方法学支持。
본문장IRT상용자료일모형의합검험통계량x2화G2인입인지진단영역，구체토론료저량개통계량재인지진단자료일모형의합검험적가행성급기정사효과，위연구자급실제응용자재인지진단자료모형의합검험중제공차감급방법학지지。
Cognitive diagnosis or diagnostic classification models （e. g. , Rupp, Templin, ＆ Henson, 2010） hold great promise for such contexts because they can potentially support relatively finegrained information about respondents that can be used for developing targeted interventions. Cognitive diagnosis enabled us to learn more about the internal psychological rules and human processing mecha nisms, and to realize the diagnostic assessment of individual cognitive developing status, which provided his/her cognitive procedure, processing skills and knowledge structures, including their advantages and deficiencies. All such information is useful for teaching and learning, and important to individual development. Recently, more and more attention has been paid to cognitive diagnosis （CD）. As with any psychometric models, the validity of inferences from cognitive diagnosis models （CDMs） determines the extent to which these models can be useful. For inferences from CDMs to be valid, it is crucial that the fit of the model to the data is ascertained. Model checking is a crucial part of any model that is based on statistical analysis and provides a vital sanity check that the theory under lying the model can actually predict the phenomena observed in the data. Model checking can also suggest improvements to the model and hence the underlying process that generated the data. This paper will approachthe topic under the cognitive diagnosis model. Under Item Response Theory （IRT） framework, the,y2 and G2 statistics are the two popular and well done goodness of fit test statistics. This paper incorporates these two statistics into the cognitive diagnosis theory. The Monte Carlo simulation method was used to investigate the feasibility and effectiveness of thestatistics in the cognitive diagnosis theory. The X2 and G2 statistics were also used to analyze the real data to detect the application of them in real work. The findings showed that：（ 1 ） During the goodness of fit test of cognitive diagnosis, the type I and II error rate of both statistics were less than 5% , which indicated that they could effectively detect the misfit items and were available in cognitive diagnosis. Furthermore, thexZstatistic works better than the G2 statistic. （2） It was found that the effectiveness of the X2 and G2 statistics would be affected by the test length, the sample size and the number of the attributes. （3） Both statistics could effectively find out the misfit items whose attributes were incorrectly calibrated. Although the results of this work are encouraging, additional work is needed to further understand the modeldata fit evaluation un der the CDM context, and to broaden the generalizability of the current findings. In cognitive diagnosis modeling, the goodnessoffit evaluation may be complicated and challenging because of the many possible causes of misfit. It often requires simultaneous consideration of the impacts of different components of the model. But, as this study shows, thex2 and G2 statistics can provide a viable means of detecting misspecifications. With additional developments, the itemlevel and testlevel fit statistics can be integrated to finetune the process of evaluating the modeldata fit in cognitive diagnosis modeling in the future.