大尺度教育测评中IRT等值方法的比较研究  

Comparison of Equating Methods in Common-item Non Equivalent Group Design Based on Item Response Theory in Large Scale Assessment

作　　者：杨涛[1] 辛涛[2] 高燕[3] 

机构地区：[1]北京师范大学教育统计与测量研究所,北京100875 [2]北京师范大学发展心理研究所,北京100875 [3]北京师范大学认知神经科学与学习国家重点实验室,北京100875

出　　处：《中国软科学》2013年第12期158-164,共7页China Soft Science

摘　　要：世界各国和国际组织都把大尺度教育测评作为关注教育质量的有效方式。有效发挥大尺度教育测评的作用离不开对其测评技术和方法的深入研究。等值是大尺度教育测评中数据处理的重要方法,有助于实现测验的公平。研究采用共同题非等组设计,利用实际数据,比较3种等值方法(同时性校准、Stocking&Lord分别校准方法和Mean-Mean分别校准方法)与2种分别校准链接方案(集中式链接、链式链接)的题目参数等值精度。结果显示:多题本共同题设计下,Stocking&Lord方法的估计精度要高于Mean-Mean方法;Mean-Mean和Stocking&Lord两种分别校准等值方法在集中式链接方案和链式链接方案中的参数估计精度都比同时性校准的精度高;在链式链接方案中,随着链接题本数量的增加,参数等值精度在下降,在集中式链接方案中等值精度则没有这样明显的规律;等值方法对精度的影响要大于链接方案对等值精度的影响。其中,不同等值方法虽然在不同研究中分别被采用过,但同时考虑不同题本链接方案在国内外均没有实证研究。此研究结果,既弥补了国内外有关多题本共同题设计下不同等值方法以及不同题本链接方案的实证研究的不足,也为我国相关大尺度教育测评的等值设计提供参考和实证依据,进而更好地实现测验的公平。In the common -item nonequivalent groups design,three equating methods were compared： Stocking ＆ Lord approach, mean -mean approach, and concurrent calibration method. Meanwhile, two linking plans were studied： the focus equating plan and the chain equating plan. The research used the real test data with large sample criteria as equa- ting criteria and RMSD as precision indicator. The results showed that, in our research situation, the stocking ＆ Lord method performed better than the Mean-Mean method for the calibration of the item discrimination and difficulty parame-ters. The two methods were better than that of concurrent calibration. The precision of item parameters turned down as the booklet number increasing in the chain equating plan, while it was not the same case in the focus equating plan. Generally speaking, the impact of equating methods was larger than that of equating plans to the equating precision.

关 键 词：IRT 等值 链接方案 

分 类 号：G40-032[文化科学—教育学原理]