Teams link https://teams.microsoft.com/meet/322711680374516?p=hr2nnC37mBds0cCtOw
Add to Google Calendar
Abstract
Non-observable constructs such as personality, intelligence, and well-being are at the core of research on human behaviour and cognition. Latent variable methods (e.g., factor analysis, structural equation modelling) are therefore an indispensable tool for research in the social and behavioural sciences. However, contemporary applications increasingly involve high-dimensional settings and structured data collections, including multigroup designs and multiblock (multi-view) measurements on the same observational units. Classical factor analysis methods are not well suited to these scenarios.
In this talk I will discuss a regularized exploratory approximate factor analysis framework that addresses these challenges. High dimensionality is handled via an approximate factor model formulation, while identifiability and estimation stability are enforced through structured regularization. Specifically, the framework relies on penalties and constraints to encourage sparse loadings (simple structure) and alignment of factors across groups or data blocks.
Short Bio
Katrijn Van Deun is a Professor in Data Science for the Social and Behavioral sciences at Tilburg University, affiliated to the department of Methodology & Statistics since 2014. She received her PhD in Psychometrics from the KU Leuven in 2005 and worked as a postdoc in computational biology at that same university. Her current research focuses on the development of latent variable methods for high-dimensional (multiview) data, to support the development of personalized multidisciplinary treatment plans in collaboration with behavioral and medical scientists. She is an elected member of the International Statistical Institute and recipient of a prestigious personal Vici grant from the Dutch Research Council.
More information at News U. Porto