Statistical models frequently employ nonlinear effects, interactions, and multiresolution basis expansions to represent complex relationships between predictors and responses. Determining the appropriate level of model complexity, however, remains a fundamental challenge. This paper introduces \textit{Inherited Local Shrinkage} (ILS), a Bayesian global-local shrinkage framework for adaptive functional form discovery that incorporates strong heredity among hierarchically related predictors. ILS propagates shrinkage through directed acyclic graphs (DAGs) encoding relationships among predictors, allowing increasingly complex terms to contribute only when supported by their simpler ancestors. This construction provides a common prior framework for a broad class of hierarchical regression models. We demonstrate its application to polynomial regression, as well as multiresolution modeling with two-dimensional Haar wavelet and eigenvalue-ordered spline (EVOS) bases. Although these examples illustrate specific basis systems, ILS is not tied to any particular basis construction and can be applied to any regression model in which relationships among predictors can be represented by a DAG, regardless of the graph's size or topology. Simulation studies demonstrate that ILS achieves competitive predictive performance relative to existing heredity-constrained methods while recovering underlying functional structure. Spatial applications further illustrate its ability to adaptively regulate model complexity across multiple resolutions, identifying where finer-scale detail is supported by the data. Together, these results establish ILS as a flexible framework for hierarchical complexity control and adaptive functional form discovery in Bayesian regression.

翻译：暂无翻译