Online Program Home
My Program

Abstract Details

Activity Number: 165 - SPEED: Environmetrics: Spatio-Temporal and Other Models
Type: Contributed
Date/Time: Monday, July 30, 2018 : 10:30 AM to 12:20 PM
Sponsor: Section on Statistics and the Environment
Abstract #329857 Presentation
Title: A Bayesian Approach to Trend Filtering for Spatially Confounded Data
Author(s): Adam Walder*
Keywords: Spatial Statistics; Trend Filtering; Bayesian ; Random Field

Trend filtering proposed by Kim et. al. (2009), provides a computationally feasible tool for performing nonparametric regression resembling splines in the univariate setting. Wang et al. (2011) provided an extension from the univariate case to graphs with their proposed Graph Trend Filtering. In our work, we propose a continuous-space Bayesian version of Graph Trend Filtering. Aberg and Podgorski (2011) provide motivation for implementing non-stationary second order random fields with heavier tails to provide models with better flexibility. To provide a model that adapts better to local behavior, we propose a model that fits a spatial term to a Laplace Random Field. We illustrate the utility of this approach on a study of spatial disease mapping, where we show that our proposed approach is less prone to spatial confounding.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2018 program