关键词: GIRT模型, 矩阵抽样, 群体水平项目参数, 学校能力

Abstract:

Traditionally the assessment of group is based on the assessment of individual, which is the mean of them. Now a new method, group-level IRT, which had more advantages over IRT, can realize the assessment of group without the individual assessment. This paper tried to apply the GIRT model into the assessment on English reading comprehension ability about school. Findings showed: All school abilities were almost within the internal [-1,1], there were no extreme ability values. When all schools were classified into three types according to the rules of this province, the mean of school abilities of them were different significantly. For school, almost all items were easy and had moderate discrimination. For students, all items were also easy, but they had good discrimination as a whole. Furthermore, the assessment consequences of school between GIRT and IRT were significant correlated, and they were also significant correlated with the mean scores of school. It could verified that the item parameters of student-level and group-level were significant correlated, which identified that the function relationships between the items parameters of GIRT and IRT model were existed. All these mean that GIRT model was a feasible group assessment method in practice.

Key words: group-level IRT model, matrix sample, group-level item parameters, school ability