A1345
Title: HFG LASSO: A method for models with hierarchical categorical predictors
Authors: Nikolas Krstic - University of British Columbia (Canada) [presenting]
Gabriela Cohen Freue - University of British Columbia (Canada)
Abstract: Hierarchical structures of Categorical Predictors are characterized by the levels of some Predictors being nested within levels of others, forming a Hierarchical tree. These structures are commonly encountered in disciplines like biology, business, and public health. Such sets of Categorical Predictors are referred to as Hierarchical Categorical Predictors (HCPs). Given that HCPs often have high cardinality at the most granular layer, their inclusion within linear regression models can increase model complexity substantially. The Hierarchical Fused Group LASSO (HFG LASSO) penalty is proposed for linear regression models. This penalty fuses together an HCP's coefficients based on the groups defined by its underlying Hierarchical tree, both promoting model sparsity and fusing coefficients according to their Hierarchical groups. A corresponding fitting algorithm and its implementation are also presented. The performance of HFG LASSO with respect to prediction, coefficient estimation, and Group recovery is assessed on both simulated data and an applied dataset. The results demonstrate that HFG LASSO excels at Group recovery and coefficient estimation compared to other candidate methods, while maintaining comparable predictive performance. Although the focus is on applying HFG LASSO to HCPs, the method is generalizable to other Predictors forming Hierarchical structures.
