項目反応理論による多値データの分析について : 段階反応モデルと部分得点モデル

概要

論文の詳細を見る
Generally, item response models, including the two-parameter logistic model, can only deal with binary data, such as correct or incorrect responses－not polytomous data. In item response theory, polytomous statistics are obtained using special models such as Samejima's graded response model. However, since these models are complex, it is not easy to estimate their parameters. In contrast, Fujimori's partial test score model makes estimating and interpreting parameters easy, because this model is a natural expansion of the two-parameter logistic model. In the present study, the partial test score model having these features is compared to the graded response model, which is used widely as an analysis model for polytomous data. The comparison was performed through simulations and analysis of real data. The partial test score model yielded good results in terms of AIC（Akaike's information criterion）, as well as model reproducibility. It also showed good results in terms of estimation of parameters. That is, in the case of the graded response model, bias was observed in estimators when the number of items was small, whereas no such biases were observed for the partial test score model, which is indicative of the advantage of using the proposed model.
文教大学の論文