基于双因子模型的测验总分和维度分的合成方法

发布时间：2018-11-10 11:17

【摘要】：双因子模型可以同时包含一个全局因子和多个局部因子,在描述多维测验结构时有其独特优势,近些年应用越来越广泛。文章基于双因子模型,提出了4种合成总分和维度分的方法,分别是:原始分法,加和法,全局题目加权加和法和局部题目加权加和法,并采用模拟的方法,在样本量、测验长度、维度间相关变化的条件下考察了这些方法与传统多维IRT方法的表现。最后,通过实证研究对结果进行了验证。结果显示:(1)全局加权加和法和局部加权加和法,尤其是局部加权加和法合成的总分和维度分与真值最接近、信度最高。(2)在维度间相关较高,测验长度较长的条件下,局部加权加和法的结果较好,部分条件下甚至优于多维IRT法。(3)仅有局部加权加和法合成的维度分能够反应维度间真实的相关关系。
[Abstract]:The two-factor model can include both a global factor and a number of local factors. It has its unique advantages in describing the multi-dimensional test structure and has been used more and more widely in recent years. Based on the two-factor model, four methods of synthesizing total score and dimension score are proposed in this paper. They are the original method, the addition method, the global topic weighted addition method and the local topic weighted sum method, and the simulation method is used in the sample size. The performance of these methods and the traditional multidimensional IRT method was investigated under the condition of the variation of the test length and the correlation between dimensions. Finally, the results are verified by empirical research. The results showed that: (1) the total score and dimension score of global weighted sum method and local weighted summation method, especially the local weighted summation method, were the closest to the true value, and the reliability was the highest. (2) the correlation between the two dimensions was higher. When the length of the test is longer, the results of the local weighted summation method are better than that of the multidimensional IRT method. (3) only the dimension points synthesized by the local weighted addition method can reflect the true correlation between the dimensions.
【作者单位】：北京师范大学心理学部;
【基金】：国家自然科学基金(31571152) 北京市与中央在京高校共建项目(019-105812) 未来教育高精尖创新中心、中央高校基本科研业务费专项资金资助
【分类号】：B841.7

【相似文献】