CMStatistics 2023: Start Registration
View Submission - CFE
A1091
Title: Spike-and-slab group Dirichlet-Laplace priors for sparse shrinkages Authors:  Deborah Gefang - University of Leicester (United Kingdom) [presenting]
Abstract: Dirichlet-Laplace (DL) priors of prior studies have proved powerful in variable selection and parameter estimations. However, similar to many other popular global-local shrinkage priors, DL priors usually produce posterior means (medians) that are close to zero but not exactly zero for true zero parameters, and underestimate the magnitudes of true non-zero parameters when the number of variables is large. Spike-and-slab group Dirichlet-Laplace is introduced priors to identify variable groups and sparsity both at the group level and within groups.