多面Rasch模型在主观题评分培训中的应用  

Application of Many-Facet Rasch Model in Rater Rraining for Subjective Items

作　　者：李中权[1] 孙晓敏[1] 张厚粲[1] 张立松[1] 

机构地区：[1]北京师范大学

出　　处：《中国考试》2008年第1期26-31,共6页journal of China Examinations

摘　　要：主观题的评分受到很多因素的影响，如评分者的知识水平、综合能力和个人偏好等。这些评分者偏差不仅会导致不同评分者之M存在主观差异，也会到导致同一评分者在不同的时间也具有主观不稳定性。最终导致主观题评分信度的降低。本研究将多面Rasch模型运用到某国家级考试论述题的评分培训中。通过分析6名有经验评分者对58份试卷的试评数据，鉴别出四种评分者偏差，然后据此对每个评分者进行个别反馈,从而提高评分的客观性和精确性。Rating of Subjective items may be influenced by many factors, such as knowledge, ability and preference of raters. They result in not only inter-rater inconsistence, but also intra-rater instability, both of which harm to the rating reliability for subjective items. In the presents study, many-facet Rasch model was applied in rater training program for a national examination. Four types of rater biases were identified through data analysis of 6 raters on an essay item for 58 examinees. Then raters were received individual feedbacks on rating behavior to improve rating objectivity and accuracy.

关 键 词：多面RASCH模型 主观评分 评分者培训 

分 类 号：G405[文化科学—教育学原理]