Work

Reduced varying coefficient models for regional quantile regression with multiple responses

Year: 2026

Type: article

Abstract: Analyzing multiple outcome variables via regional quantile regression in high-dimensional settings poses significant statistical and computational challenges. In this paper, we propose a new framework that models multivariate quantile varying coefficients using principal component functions, enforcing a low-rank structure on the coefficient matrix to achieve parsimony and interpretability. Our approach augments this representation with a KNN-fused LASSO penalty to capture shared dynamic patterns and identify latent clusters within the principal components. Through comprehensive simulation studies, we demonstrate that our method consistently provides accurate estimates and robust performance under various high-dimensional scenarios. We further illustrate its practical utility with two real-world health datasets, where our approach uncovers complex, quantile-specific associations between predictors and multiple correlated outcomes across a time index.

Source: Biometrics

DOI: 10.1093/biomtc/ujag040

Authors: Woorim Jung, Seyoung Park, Hyokyoung G Hong, Eun Ryung Lee

Institutions: Sungkyunkwan University, Yonsei University, National Institutes of Health, National Cancer Institute, Division of Cancer Epidemiology and Genetics

Language: English

Topic: Statistical Methods and Inference

Subfield: Statistics and Probability

Field: Mathematics

Domain: Physical Sciences

Open Access status: green