Pengyuan (Travis) Yang

Methodology and Statistics
Tilburg School of Social and Behavioral Sciences
Tilburg University

Email

Project
Embracing the complexity of contemporary research: The next generation of structural equation modelling tools

Latent variable methods are an important tool for social and behavioral research in terms of exploring non-observable constructs. Item factor analysis (IFA), as a latent variable method, specifically deals with binary/polytomous item responses data by making use of both factor analysis and item response theory (IRT). Nevertheless, the application of IFA in a high-dimensional setting has not been systematically developed. This project will focus on developing the IFA method to solve high-dimensional and complex data problems. Specifically, joint maximum likelihood estimation will be used together with regularization (see also Sun et al., 2016 & Chen et al., 2019). This approach avoids the problems encountered by traditional IFA approaches in high-dimensional settings. It also allows to build on the recent developments in optimization (e.g alternating optimization framework; see Huang, Sidiropoulos, & Liavas, 2016, Boyd et al., 2011). Based on this approach, measurement models for multiblock/multigroup data, personalized measurement models and structural models will be developed for comprehensive application.

Supervisors
Prof. Dr. Katrijn van Deun
Dr. Wilco Emons

Financed by
NWO Vici grant

Period
1 September 2024 – 1 September 2028