|
Activity Number:
|
84
- Advances in Spatio-Temporal Statistics with Applications to Environmental Data
|
|
Type:
|
Topic-Contributed
|
|
Date/Time:
|
Monday, August 9, 2021 : 10:00 AM to 11:50 AM
|
|
Sponsor:
|
Section on Statistics and the Environment
|
|
Abstract #317484
|
|
|
Title:
|
Scalable Forward Sampler Backward Smoother Based on the Vecchia Approximation
|
|
Author(s):
|
Marcin Jurek* and Matthias Katzfuss and Pulong Ma
|
|
Companies:
|
University of Texas and Texas A&M University and Duke University / SAMSI
|
|
Keywords:
|
state-space model;
smoothing;
filtering;
spatio-temporal statistics;
data assimilation;
remote sensing
|
|
Abstract:
|
We propose an approximation to the Forward Filter Backward (FFBS) sampler commonly used in Bayesian statistics when working with linear Gaussian state-space models. Traditional FFBS has proved immensely useful and fast but it requires inverting covariance matrices that have the size of the latent state vector. The computational burden associated with this operation effectively prohibits its applications in high-dimensional settings. In this paper, we propose an approach based on the hierarchical Vecchia approximation of Gaussian processes, successfully used in spatial statistics. We extend the Vecchia approximation to variables without a spatial reference and use correlation distance instead. This allows us to apply the scalable version of the FFBS to any type of Gaussian data.
|
Authors who are presenting talks have a * after their name.
