A0489
Title: Varying coefficient-based graphical regression for estimating undirected graphs
Authors: Zeya Wang - University of Kentucky (United States) [presenting]
Abstract: The purpose is to discuss regression-based approaches for estimating undirected graphical models, where the goal is to model conditional dependencies among random variables as functions of subject-level covariates. These frameworks enable the incorporation of observation-specific heterogeneity into network estimation, allowing the learned graph structures to vary with subject-level characteristics. A key class of these models is built on varying coefficient foundations, providing a flexible and principled framework for capturing how aspects of graph topology, such as the presence or strength of edges, depend on covariates, typically through linear relationships. A brief overview of recent advances is provided in this area. Building on this foundation, the focus is on the aspects that accommodate nonlinear relationships between covariates and edge strengths, aiming to more accurately capture complex variations in network structure.
