ISSN 1671-3710
CN 11-4766/R
主办:中国科学院心理研究所
出版:科学出版社

心理科学进展 ›› 2018, Vol. 26 ›› Issue (2): 358-367.doi: 10.3724/SP.J.1042.2018.00358

• 研究方法 • 上一篇    下一篇

 全信息项目双因子分析:模型、参数估计及其应用

毛秀珍1;夏梦连2;辛 涛3   

  1.  (1四川师范大学教育科学学院; 2四川师范大学教师教育与心理学院, 成都 610068) (3北京师范大学协同创新中心, 北京 100875)
  • 收稿日期:2017-06-09 出版日期:2018-02-15 发布日期:2017-12-26
  • 通讯作者: 辛涛, E-mail: xintao@bnu.edu.cn
  • 基金资助:
     国家自然科学基金青年项目:多维计算机化自适应测验选题策略的研究与应用(31400897); 多维项目反应理论的改进及应用(11201313)。

 Full-information item bifactor analysis: Model, parameter estimation and application

 MAO Xiuzhen1; XIA Menglian2; XIN Tao3   

  1.  (1 School of Educational Science, Sichuan Normal University; 2 School of Teacher Education and Psychology, Sichuan Normal University, Chengdu 610068, China) (3 Collaborative Innovation center, Beijing Normal University, Beijing 100875, China)
  • Received:2017-06-09 Online:2018-02-15 Published:2017-12-26
  • Contact: XIN Tao, E-mail: xintao@bnu.edu.cn
  • Supported by:
     

摘要:  全信息项目双因子分析作为一种重要的统计方法, 使得双因子模型在近20年得到重新认识和广泛应用。首先详细介绍了全信息项目双因子分析方法的概念、特征、模型基础以及参数估计中体现的维度缩减思想, 然后例举全信息项目双因子分析在分析测验结构、分数解释和计算机化自适应测验中的应用。全信息项目双因子分析中双因子模型符合大量心理、教育与医学测验的结构特征, 其维度缩减方法能显著降低计算量, 因而具有广阔的应用前景。结合当前研究现状对全信息项目双因子分析的相关研究, 如:参数估计、模型特征、拟合检验、量表连接、项目功能差异及其在计算机化自适应测验中的应用提出一些思考和建议。

关键词: 双因子模型, 全信息项目双因子分析, 双因子项目反应理论模型, 维度缩减

Abstract:  Full-information item bifactor analysis is an important statistical method in psychological and educational measurement, which can be seen as a rediscovery of the classical bifactor model and has seen wide applications in the past two decades. The item response model of full-information item bifactor analysis is described upon the introduction of the conception and characterization of full-information item bifactor analysis. Further, we introduced the dimension reduction method used in parameter estimation. Then, examples are provided for applications of the full-information item bifactor model in test measurement structure exploration or confirmation, score interpretation, and computerized adaptive testing. The measurement structure of full-information item bifactor analysis is accordance with most tests in the areas of psychology, education, and medical science. With the advantage in dimension reduction, it is believed that the full-information item bifactor analysis could be valuable and useful in various situations. At last, some future research directions and suggestions are put forward including parameter estimation, test linking, differential item functioning, model fit testing, and application of bifactor item response theory to computerized adaptive testing.

Key words:  bifactor model, full-information item bifactor analysis, bifactor item reponse theory model, dimension reduction

中图分类号: