Next-generation factor models for social data science

Reference No.	2026b003
Type/Category	Grant for International Project Research-Workshop (I)
Title of Research Project	Next-generation factor models for social data science
Principal Investigator	Gabriel Wallin（School of Mathematical Sciences - Lancaster University・Lecturer/Assistant Professor）
Research Period
Keyword(s) of Research Fields	Factor analysis; Latent variable modeling; Probabilistic dimension reduction; Rotational indeterminacy; Computational algebra; Statistical methodology
Abstract for Research Report	The proposed workshop aims to bring together researchers working on factor analysis and latent variable modeling to discuss recent methodological advances, emerging challenges, and promising interdisciplinary directions. Factor analysis is a cornerstone of modern statistical methodology, underpinning probabilistic dimension reduction techniques that are widely used across the social, behavioural and health sciences, as well as in machine learning. Despite its long history, fundamental challenges remain, particularly regarding interpretability, identifiability, and computational feasibility. A central motivation for this workshop is the problem of rotational indeterminacy in factor analysis. While rotation techniques are routinely applied to achieve interpretable solutions, existing approaches often rely on heuristic optimization methods that are sensitive to initialization and may converge to suboptimal solutions. Recent work by Fukasaku et al. (2025a, 2025b) has demonstrated how tools from computational algebra can be used to characterize all stationary points of rotation criteria in orthogonal factor models, offering an alternative to traditional numerical optimization. The current collaborative research with the applicant (Dr. Gabriel Wallin from Lancaster University) extends this perspective to oblique factor models, which allow correlated latent factors and are more realistic in many applied settings. Beyond this specific line of work, the workshop is designed to showcase state-of-the-art research in latent variable modeling more broadly. Topics will include generalized linear latent variable models, spatial and multiresolution extensions of factor models, and change-point detection in sequential data with latent structure. By bringing together researchers from statistics, applied mathematics, and related fields, the workshop aims to highlight how methodological innovation often arises at the interface of disciplines. The expected outcomes of the workshop include: (i) dissemination of cutting-edge research to a diverse audience, (ii) strengthening existing collaborations between Kyushu University and the School of Mathematical Sciences at Lancaster University, (iii) fostering new interdisciplinary connections, particularly between statistics and applied mathematics, and (iv) identifying open research questions that can guide future collaborative projects.
Organizing Committee Members (Workshop) Participants (Short-term Joint Usage)	Dr. Gabriel Wallin（School of Mathematical Sciences, Lancaster University, UK・Lecturer (Assistant Professor)） Dr. Ryoya Fukusaku（Faculty of Mathematics, Kyushu University・Assistant Professor） Professor Kei Hirose（Institute of Mathematics for Industry, Kyushu University・Professor）