项目名称: 多维项目反应理论的改进及应用
项目编号: No.11201313
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 数理科学和化学
项目作者: 付志慧
作者单位: 沈阳师范大学
项目金额: 22万元
中文摘要: 与经典教育测量方法相比,基于项目反应理论的教育统计与心理测量技术呈现出愈来愈多的优势。多维项目反应理论(MIRT)是基于因子分析和单维项目反应理论两大背景下发展起来的一种新型测验理论。这是个需要重点研究的领域, 因为目前使用的大多数教育测验都是评估多种特质的。然而基于MIRT的很多理论有待于继续研究和完善.本项目集中解决如下三个关键问题::⑴提出新的多维度多级评分能力评价模型。(2)MIRT参数估计程序的开发。基于数据扩充技术,利用MCMC算法,对所建模型提出相应的Gibbs抽样估计方法,并开发出相应的应用软件。在此基础上,针对缺失反应数据问题, 采用变量分离法讨论其缺失机制,并给出缺失数据处理方法。(3)对纵向反应数据建模。在取得的理论研究成果基础上,配套开发基于MIRT的参数估计软件, 并应用于我国大型测验中,为相关职能部门决策提供科学依据.
中文关键词: 缺失数据;多层高阶项目反应模型;四参数项目反应模型;贝叶斯估计方法;Copula
英文摘要: Compared to the classical test theory,the educational satistics and psychological measurement based on the item response theory presents more and more advantages.Multidimensional Item Response Theory (MIRT) is a new test theory, which develops on the basis of factor analysis and unidimensional IRT.However, many MIRT-based theories still remain to be explored at continuing and be perfected.This project focused on solving the following three problems:(1)Propose new multidimensional and polytomous credit ability evaluation model.(2)Develop the paremeter estimating program for the MIRT model.Based on the data augmentation technology,propose the corresponding Gibbs sampling method and develop application software serving for above estimating methods.On that basis,Through the variable splitted procedure, the missing mechanism and disposal was put forward to do well in the missing data problem.(3)Modeling the longitudinal data.Based on the theoretical results we obtained,develop the associated parameter estimating software. This will be used in the large scale test,and provide science basis for the relevant functional departments.
英文关键词: Missing data;Multilevel and high-order item response model;Four-parameter item response model;Bayesian estimating method;Copula