Cognitive Diagnosis Models: Identifiability, Estimation, and Applications

认知诊断模型：可识别性、估计和应用

• 批准号: 1659328
• 负责人: Gongjun Xu
• 金额: $ 21.5万
• 依托单位: Regents of the University of Michigan - Ann Arbor
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-06-01 至 2020-05-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1659328&HistoricalAwards=false
• 关键词: Cognitive; Diagnosis; Models; Identifiability; Estimation

Cognitive diagnosis models (CDMs) are state-of-the-art psychometric models in education, psychology, and other social sciences. This research project will introduce a comprehensive theoretical and methodological framework that will make construction and analysis of CDM-based assessments more practicable. This development will provide a better understanding of the skills and cognitive processes involved in cognitive assessments. From a societal perspective, the integration of CDMs with cognitive and learning sciences will provide a powerful tool for identifying specific problems and difficulties that students encounter in skill acquisition. This interdisciplinary research project will help design a blueprint for mapping out timely, appropriate, and targeted interventions. The results of this project have the potential to positively impact STEM education and the training of students. Graduate students will be involved in the conduct of the research. Publicly available software also will be developed.This research project addresses fundamental identifiability and estimation issues of CDMs. Specifically, the project will address identifiability issues for general CDMs and provide practical guidelines for designing identifiable and statistical valid diagnosis tests. The project will address the challenging issue of Q-matrix validation and estimation. The research will develop computationally efficient methods to estimate the Q-matrix and detect possible misspecification of the Q-matrix and provide the related theoretical justification. The project also will develop an accessible computer program that can be used in conjunction with the proposed theory and methods. Extensive simulation studies will be performed to validate the performance of the estimation methods, and a variety of real data sets will be analyzed.

认知诊断模型（CDMs）是教育学、心理学和其他社会科学中最先进的心理测量模型。本研究项目将提出一个全面的理论和方法框架，使基于清洁发展机制的评估的构建和分析更加切实可行。这一发展将使人们更好地理解认知评估所涉及的技能和认知过程。从社会的角度来看，CDM与认知和学习科学的整合将提供一个强大的工具，用于识别学生在技能获取中遇到的具体问题和困难。这个跨学科的研究项目将有助于设计一个蓝图，制定及时，适当和有针对性的干预措施。该项目的成果有可能对STEM教育和学生培训产生积极影响。研究生将参与研究的进行。本研究项目涉及清洁发展机制的基本可识别性和估算问题。具体而言，该项目将解决一般CDM的可识别性问题，并为设计可识别和统计有效的诊断测试提供实用指南。该项目将解决Q矩阵验证和估计的挑战性问题。该研究将开发计算效率高的方法来估计Q-矩阵和检测可能的错误指定的Q-矩阵，并提供相关的理论证明。该项目还将开发一个可访问的计算机程序，可以与所提出的理论和方法结合使用。将进行广泛的模拟研究，以验证估计方法的性能，并将分析各种真实的数据集。

项目成果

LIKELIHOOD RATIO TEST IN MULTIVARIATE LINEAR REGRESSION: FROM LOW TO HIGH DIMENSION

• DOI: 10.5705/ss.202019.0056
• 发表时间: 2021-07-01
• 期刊: STATISTICA SINICA
• 影响因子: 1.4
• 作者: He, Yinqiu;Jiang, Tiefeng;Xu, Gongjun
• 通讯作者: Xu, Gongjun

Transformed Dynamic Quantile Regression on Censored Data

截尾数据的变换动态分位数回归

• DOI: 10.1080/01621459.2019.1695623
• 发表时间: 2020
• 期刊: Journal of the American Statistical Association
• 影响因子: 3.7
• 作者: Chu, Chi Wing;Sit, Tony;Xu, Gongjun
• 通讯作者: Xu, Gongjun

A Regularization-Based Adaptive Test for High-Dimensional Generalized Linear Models

• 发表时间: 2020-07
• 期刊: Journal of machine learning research : JMLR
• 作者: Chong Wu;Gongjun Xu;Xiaotong Shen;W. Pan

Identifiability of Hierarchical Latent Attribute Models

• DOI: 10.5705/ss.202021.0350
• 发表时间: 2019-06
• 期刊: Statistica Sinica
• 影响因子: 1.4
• 作者: Yuqi Gu;Gongjun Xu

Gaussian variational estimation for multidimensional item response theory

多维项目响应理论的高斯变分估计

• DOI: 10.1111/bmsp.12219
• 发表时间: 2020
• 期刊: British Journal of Mathematical and Statistical Psychology
• 影响因子: 2.6
• 作者: Cho, April E.;Wang, Chun;Zhang, Xue;Xu, Gongjun
• 通讯作者: Xu, Gongjun