SUPR
Bayesian inference for locally stationary processes
Dnr:

NAISS 2024/22-829

Type:

NAISS Small Compute

Principal Investigator:

Mattias Villani

Affiliation:

Stockholms universitet

Start Date:

2024-06-05

End Date:

2025-07-01

Primary Classification:

10106: Probability Theory and Statistics

Webpage:

Allocation

Abstract

Temporal, spatial and spatiotemporal data are often nonstationary with properties changing over time. This project develops computationally efficient simulation-based methods for Bayesian learning and prediction in locally stationary processes with model for the parameter evolution following recently proposed dynamic shrinkage processes. The models will be developed both in the time and frequency domain, and a particular direction is to develop efficient models for handling time-varying seasonality.