Bayesian lasso in factorial experiment Designs with application

H. Abed, Maha; H. AL-Sharoot, Mohammed

doi:10.33916/qjae.2026.01241250

	Bayesian lasso in factorial experiment Designs with application
AL-Qadisiyah Journal For Administrative and Economic sciences
Volume 28, Issue 1, April 2026, Pages 241-250 PDF (1.78 M)
Document Type: Research Paper
DOI: 10.33916/qjae.2026.01241250
Authors
Maha H. Abed; Mohammed H. AL-Sharoot
University of AL-Qadisiya
Abstract
This paper develops the Factorial Bayesian LASSO (FBLASSO), a hierarchical shrinkage method designed for high-dimensional factorial experiment designs where numerous main and interaction effects must be estimated simultaneously. The model incorporates distinct shrinkage parameters for main and interaction terms, enabling adaptive penalization and improving estimation stability in the presence of multicollinearity. A controlled Monte Carlo simulation, reflecting realistic agricultural conditions, demonstrates that FBLASSO achieves lower mean squared error and higher true positive rates compared with Ordinary Least Squares, classical LASSO, and Ridge regression. The method is further applied to a real wheat field experiment conducted in Al-Qadisiyah Governcy, Iraq, involving fertilizer levels, irrigation regimes, and cultivar types. Results show that fertilizer and irrigation exert the strongest main effects on yield, while only a limited subset of interactions is retained by the model. These findings highlight the effectiveness of FBLASSO in recovering influential factorial effects and producing interpretable results in complex agricultural experiments.
Keywords
Bayesian LASSO; Factorial designs; Shrinkage methods; Agricultural experiments; Hierarchical modeling

Statistics Article View: 56 PDF Download: 87