语言测试研究中的多层面Rasch模型——原理简介和研究综述

Many-Facet Rasch Model in language testing research:A brief introduction and literature review

语言运用测试(language performance assessment)是各种大规模语言考试中不可或缺的部分,但因为其测试环境中不可避免地引入了评分员的主观判断、测试任务的难度、评分标准的设定和使用情况等因素,使其信度、效度及考试的公平性都受到了质疑。多层面Rasch模型是项目反应理论(IRT)中经典Rasch模型在多个维度上的延伸,它的主要优势在于可以将考试环境中多个影响考生最终得分的因素纳入同一个数学模型中进行分析,并估算出各个层面的因素对考生成绩的影响程度。本文旨在对MFRM的工作原理和基本模型进行简单介绍,并系统梳理语言测试领域运用MFRM进行的相关研究,以期让读者更好地了解如何在语言测试研究中有效地运用这种统计方法。

Language performance assessment has been an indispensible part of many large-scale language tests. How- ever, the reliability and validity of such assessment tasks are challenged due to the subjective rater judgment, differ- ent levels of task difficulty and the setting and use of rating scales introduced into the test context. Many-Facet Rasch Model （MFRM） is the muhi-Faceted extension of the classic Rasch model in Item Response Theory. The major ad- vantage of MFRM is to incorporate multiple factors which may influence the final scores of candidates into one unified mathematical model, simultaneously to analyze them and then to estimate the extent to which each factor determines the final scores. The present paper intends to give a brief introduction to the fundamental principles and mathematical models of MFRM and make a comprehensive literature review of using MFRM in the field of language testing re- search, with the aim to provide the readers with a practical guide to MFRM in language testing research.