摘要
锚题是等值中的重要因素，在非等组锚测验设计下的等值，锚题影响着等值结果．也关乎题库建设的科学性。本文基于HSK实测数据，在IRT框架下从锚题数量、锚题内容、参数特征三个方面探讨锚题对等值结果的影响。结果表明锚题数量和锚题并不是影响IRT等值的关键因素，随着锚题数量的改变．等值结果相差不大，等值误差的大小主要与锚题的参数特征有关，锚题难度和区分度高，则等值结果比较稳定。
Anchor is the most important fact in equating. In the nonequivalent groups with an anchor design, anchor affects the result of equating, and it is concerning with the construction of item bank. Basing on the data of HSK, This paper explores the influence of anchor size, anchor context and parameter specification in IRT true score equating. The result shows that anchor size and item context are not the key elements in equating. The error of equating does＇t decrease much with the increasing of anchor size; The error of equating is in relation with anchors＇parameter specification, the anchors using in IRT equating with high difficulty and discrimination, the more robust equating result will be gotten.
出处
《中国考试》
2010年第8期9-13,共5页
journal of China Examinations
关键词
锚题数量
参数特征
题型
等值
Anchor Size
Parameter Specification
Item Context
Equating
