Factor Analysis of Multitype Recurrent Events with Applications to Modern Educational Assessment

• Activity Number: 394 - Recent Advances in Cognitive Diagnosis Modeling
• Type: Topic Contributed
• Date/Time: Tuesday, July 31, 2018 : 2:00 PM to 3:50 PM
• Sponsor: Mental Health Statistics Section
• Author(s): Yunxiao Chen*
• Companies: Emory University
• Keywords: Event history analysis; Cox model; Factor analysis; Recurrent event; Graphical model; Problem solving process

Abstract

Latent factor models have been the statistical foundation for the modern measurement theory, receiving wide applications in educational testing, personality assessment, and mental health diagnosis. With the advances in technology, innovative computer-based instruments, such as simulation-based and game-based assessments, are developed to uncover the underlying characteristics of individuals (e.g., complex problem-solving ability). Such new measurement tools collect individuals' entire process of solving one or multiple problems, producing data that cannot be handled by the classical latent factor models. In this talk, we propose a model for the factor analysis of multivariate recurrent event processes, a type of data that is commonly collected in the innovative computer-based assessments. This model contains a latent variable component and a sparse graphical component, where the former captures the underlying factors shared by the counting processes of multiple event types and the latter accounts for ad-hoc event-type specific dependence. The model is applied to analyzing data from the Programme for International Student Assessment (PISA) 2012.