B0381
Title: Trend filtering for temporal-spatial models
Authors: Daren Wang - University of Notre Dame (United States) [presenting]
Carlos Misael Madrid - University of Notre Dame (United States)
Oscar Hernan Madrid - University of California Los Angeles (United States)
Abstract: Temporal-spatial models play a crucial role in many scientific disciplines such as environmental science, epidemiology, and economics, where understanding spatiotemporal relationships is essential for accurate statistical prediction and inference. The problem of estimating the non-parametric regression function from data that exhibit both temporal and spatial dependence is tackled. A computationally efficient trend is introduced, a filtering estimator designed to handle multivariate non-parametric regression settings with arbitrary degrees of smoothness. Through matching lower bounds, the minimax optimality of the estimator is established. In addition, the analysis reveals an interesting phase transition phenomenon previously unexplored in trend-filtering literature. Simulation studies and real data applications illustrate the promising performance of the proposed approach compared to the existing methods in the literature.
