CAREER: Identifiability and Inferences for Structured Latent Attribute Models

本研究项目将解决结构化潜在属性模型(SLAMS)中的基本统计问题。这些模型是在教育、心理学以及其他社会和行为科学中开发基于诊断的评估的基础。这一职业奖将推动这些模型的理论和计算发展。该项目还将有助于更好地理解认知评估和学习科学中涉及的认知过程。从社会的角度来看，新的方法将被应用于各种教育研究。该项目将开发一种基于诊断的学习工具，以确定学生在STEM领域遇到的具体问题和困难，并向学生提供有用的反馈。研究人员将与密歇根大学自然历史博物馆合作，参与K-12教育拓展，参与科学交流研究员计划和明日科学计划等活动。本科生和研究生都将参与这项研究。随着日益丰富的诊断数据集的创建，对现有的SLAM理论和技术提出了巨大的挑战。这个项目将集中在一些公开的问题上。首先，该项目将解决SLAM的基本可识别性问题。将开发一个新的理论框架作为脚手架，以显示SLAMS的可识别结果，这将为未来的诊断设计提供实用指导。其次，该项目将开发新的程序，以解决在高维潜在属性的SLAMS中估计潜在结构的挑战。第三，该项目将为具有高维潜在属性和观测协变量的SLAM开发强大而稳健的统计推断程序。除了理论和方法的发展，该项目还将为实践者提供一种计算工具来探索和扩大SLAM的使用。这一奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

This research project will address fundamental statistical issues in Structured Latent Attribute Models (SLAMs). These models serve as a basis for developing diagnostic-based assessments in education, psychology, and other social and behavioral sciences. This CAREER award will advance the theoretical and computational development of these models. The project also will contribute to improved understanding of cognitive processes involved in cognitive assessments and learning sciences. From a societal perspective, the new methods will be applied to various educational studies. The project will develop a diagnostic-based learning tool to identify specific problems and difficulties that students encounter in STEM domains and provide students with useful feedback. The investigator will engage in K-12 educational outreach in collaboration with the University of Michigan's Museum of Natural History, participating in activities such as the Science Communication Fellows Program and the Science for Tomorrow Program. Both undergraduate and graduate students will be involved in the conduct of this research. Publicly available software also will be developed.With the creation of increasingly rich diagnostic datasets, great challenges are posed on existing SLAM theories and techniques. This project will focus on a number of open questions. First, the project will address the fundamental identifiability issue of SLAMs. A new theoretical framework will be developed as a scaffold to show the identifiability results for SLAMs, which will provide practical guidelines for future diagnostic designs. Second, the project will develop novel procedures to address the challenge in the estimation of the latent structures in SLAMs with high-dimensional latent attributes. Third, the project will develop powerful and robust statistical inference procedures for SLAMs with high-dimensional latent attributes and observed covariates. In addition to theoretical and methodological developments, the project will provide practitioners with a computational tool to explore and expand the use of SLAMs.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.