Loading...
Honey, I shrunk the irrelevant effects!: Simple and flexible approximate Bayesian regularization
Karimova,Diana van Erp,Sara Leenders,Roger Th. A. J. Mulder,Joris
Karimova,Diana
van Erp,Sara
Leenders,Roger Th. A. J.
Mulder,Joris
Abstract
In the social and behavioral sciences and related fields, statistical models are becoming increasingly complex with more parameters to explain intricate dependency structures among larger sets of variables. Regularization techniques, like penalized regression, help identify key parameters by shrinking negligible effects to zero, resulting in parsimonious solutions with strong predictive performance. This paper introduces a simple and flexible approximate Bayesian regularization (ABR) procedure, combining a Gaussian approximation of the likelihood with a Bayesian shrinkage prior to obtain a regularized posterior. Parsimonious (interpretable) solutions are obtained by taking the posterior modes. Parameter uncertainty is quantified using the full posterior. Implemented in the R package shrinkem, the method is evaluated in synthetic and empirical applications. Its flexibility is demonstrated across various models, including linear regression, relational event models, mediation analysis, factor analysis, and Gaussian graphical models.
Description
Date
2025-08
Journal Title
Journal ISSN
Volume Title
Publisher
Files
Research Projects
Organizational Units
Journal Issue
Keywords
Bayesian analysis, Penalized regression, Psychological modeling, Statistical regularization, Variable selection
Citation
Karimova, D, van Erp, S, Leenders, R T A J & Mulder, J 2025, 'Honey, I shrunk the irrelevant effects! Simple and flexible approximate Bayesian regularization', Journal of Mathematical Psychology, vol. 126, 102925. https://doi.org/10.1016/j.jmp.2025.102925