IRT与MIRT在测验垂直等值中的应用  

The Application of IRT and MIRT in Test Vertical Equating

作　　者：王怡[1] 唐文清[1,2] 刘晶[1] 张敏强[1] 李明[3] 黎光明[1] 

机构地区：[1]华南师范大学心理学院/心理应用研究中心,广州510631 [2]广西大学教育学院,南宁530004 [3]佛山市教育局教研室,佛山528000

出　　处：《心理科学进展》2014年第5期881-888,共8页Advances in Psychological Science

基　　金：全国教育科学"十二五"规划教育部重点课题(GFA111009)

摘　　要：测验垂直等值是指将测试同一心理特质的不同水平的测验转换到同一个分数量尺上的过程。IRT与MIRT是实现垂直等值的主要方法。IRT无需假设被试的能力分布,参数估计不依赖于样本,是构建垂直量表的有效方法,但测验不满足单维假设时其应用受到限制。MIRT结合IRT和因素分析的特点对IRT进行了拓展,可更有效估计多维测验的项目参数和被试能力参数,在垂直等值中有重要应用。已有研究主要探讨IRT和MIRT在垂直等值应用中的适用性、标定方法和参数估计方法,比较研究两种方法的特性。未来研究应纳入更多变量条件进行比较研究,拓展方法的应用。Test vertical equating refers to transforming the performance of students in tests of various levels to the same scale. Item Response Theory （IRT） and Multidimensional Item Response Theory （MIRT） are the main methods used in vertical equating. IRT method is an effective way to build a vertical scale because IRT models impose no requirement on the distribution of test-takers＇ ability and that item parameters are independent of the samples. However, its application is restricted when the test dose not satisfy the unidimensional assumption. MIRT is a combination of IRT and factor analysis, making it an important tool in vertical equating to estimate ability parameter and item parameter precisely in multidimensional test. Current researches focus on the applicability of IRT and MIRT, calibration method and parameter estimation method in vertical equating and comparison study of the two methods. Future studies should concentrate on including more variables when comparing IRT and MIRT in order to expand their applications.

关 键 词：IRT MIRT 垂直等值 

分 类 号：B841[哲学宗教—基础心理学]