基于多层面Rasch模型的概要写作效度验证  

A many-facet Rasch model analysis of summary writing

作　　者：李久亮[1] LI Jiuliang

机构地区：[1]北京服装学院文理学院,北京100029

出　　处：《外语测试与教学》2022年第3期38-48,共11页Foreign Language Testing and Teaching

基　　金：北京服装学院教育教学改革立项项目“中国英语能力等级量表下的大学英语写作教学与评价研究”(项目编号JG-1923)。

摘　　要：本文运用多层面Rasch模型对概要写作作为一项综合技能考试题型开展效度研究,从任务难度、评分量表、考生能力和评分员严厉度这四个方面进行了验证。被试为来自某高校的83名本科生,需要撰写两篇英文文章的英文概要,由三位评分员参与评分。结果发现,记叙文概要写作难于说明文。评分量表能够较好地区分考生的能力,但仍需进一步改进;评分单项中语言使用(LU)难度最大,原文信息使用(SU)难度最小。模型各层面内个体的拟合值基本都处于可接受的范围,其中仅有一名考生出现了非拟合的情况。评分员严厉度差异明显,但评分员内部一致性较好;偏差分析说明评分员对个别考生过于严厉或者宽松,但这些并没有显著影响测试分数。This study attempts to validate summary writing as an integrated task with a many-facet Rasch model.Facets modeled in the analysis were task,rating scale,examinee,and rater.The participants in this study comprised 83 Chinese undergraduate students,who wrote summaries for two English articles,and three trained raters.The results show that the narrative text was more difficult to summarize than the expository text.The components of the scoring scale functioned as expected by and large,despite the needs of further refinement.Among the four rating categories,Language Use is the most difficult while Source Use the easiest.The fit statistics of most facets elements fell within the acceptable range,and only one student s fit statistic fell outside of the acceptable range.The raters exercised different levels of severity but with good consistency.Bias analysis indicates that they were consistently severe or lenient with several students,but the bias had only negligible impact on examinees scores.

关 键 词：多层面RASCH模型 概要写作 效度验证 

分 类 号：H319[语言文字—英语]