A0634
Title: A Bayesian nonparametric framework for dynamic item-response theory
Authors: Maria De Iorio - National University of Singapore (Singapore) [presenting]
Abstract: Item-response theory (IRT) is widely used for the statistical analysis of questionnaire data, allowing for the differentiation of respondent profiles and the characterization of questionnaire items through interpretable parameters. A Bayesian semiparametric extension of IRT is proposed that introduces temporal dependence across repeated questionnaire administrations, accommodates repeated measurements, and jointly models responses from related subject groups (e.g., mothers and children) to enable information sharing across hierarchies. The framework further incorporates covariate information, allows for the joint modeling of questionnaire data with other longitudinal markers, and supports clustering of subjects based on their latent response profiles. The approach is built on Bayesian nonparametric priors: The Dirichlet process and the normalized generalized gamma process, facilitating the identification of clinically meaningful subgroups. The utility of the proposed methodology is demonstrated through the analysis of longitudinal psychometric questionnaire data from the Singaporean GUSTO cohort study. These data are collected from mothers and their children, aiming to investigate how various factors influence growth trajectories, developmental outcomes, and mental health.
