项目名称: 心理与教育测量中项目反应时间数据的统计建模及其应用
项目编号: No.11501094
项目类型: 青年科学基金项目
立项/批准年度: 2016
项目学科: 数理科学和化学
项目作者: 孟祥斌
作者单位: 东北师范大学
项目金额: 18万元
中文摘要: 近年来,随着计算机化考试的出现,被试项目反应时间(即答题时间)数据的记录变得异常简单。那么,为了充分利用这部分数据为心理和教育测量服务,首要任务是建立科学的统计模型,并给出相应的统计推断方法。针对现有研究的一些不足和局限,本项目拟进行以下四个方面的研究:(1) 运用生存分析中的建模方法提出更加一般化的项目反应时间模型,为项目反应时间数据的实际应用提供更为有效的分析工具。(2) 提出多级评分项目反应变量与反应时间变量的局部联合建模方法,突破局部独立性假设的限制,实现在更加一般的测试条件下被试答题速度与准确性关系的建模。(3) 运用项目反应时间的变点识别技术,提出计算机自适应测试的项目质量序贯监控方法。(4) 运用本项目提出的统计模型和相关统计推断方法,基于项目反应时间的信息设计一个更加科学的计算机自适应考试选题系统。
中文关键词: 统计建模;马尔可夫链蒙特卡罗算法;贝叶斯模型选择;项目反应理论;生存分析
英文摘要: In recent years, with the advent of the computerized testing, the recording of the item response times has become straightforward. To make full use of the response time dataset in educational and psychological measurement, the primary work is to develop the statistical models and give the necessary methods of statistical inference. Based on some deficiencies and limitations of the current studies, this project plans to solve the following four problems: (1) The more general modeling approach for response times is proposed by using the statistical methods in the survival analysis, which is a more effective tool for the actual use of the response times. (2) Develop a local joint modeling approach for the polytomous item response variable and response time variable, which does not require the assumption of conditional independence and can explain the relationship between speed and accuracy in the more general test conditions. (3) Use the change-point detection technique to develop a sequential procedure for monitoring compromised items in the item pool of computerized adaptive testing (CAT) system. (4) Based on the statistical model and the statistical inference methods proposed in this project, use the information of item response times to propose a more scientific item-selection method for CAT.
英文关键词: statistical modeling;Markov chain Monte Carlo algorithm;Bayesian model selection ;item response theory;survival analysis