A Class of Convolution-Based Models for Spatio-Temporal Processes with Non-Separable Covariance Structure

Alexandre Rodrigues, Peter J. Diggle

Research output: Contribution to journalArticle

Abstract

In this article, we propose a new parametric family of models for real-valued spatio-temporal stochastic processes S(x, t) and show how low-rank approximations can be used to overcome the computational problems that arise in fitting the proposed class of models to large datasets. Separable covariance models, in which the spatio-temporal covariance function of S(x, t) factorizes into a product of purely spatial and purely temporal functions, are often used as a convenient working assumption but are too inflexible to cover the range of covariance structures encountered in applications. We define positive and negative non-separability and show that in our proposed family we can capture positive, zero and negative non-separability by varying the value of a single parameter.

Original languageEnglish (US)
Pages (from-to)553-567
Number of pages15
JournalScandinavian Journal of Statistics
Volume37
Issue number4
DOIs
StatePublished - Dec 2010
Externally publishedYes

Fingerprint

Spatio-temporal Process
Nonseparable
Covariance Structure
Convolution
Low-rank Approximation
Covariance Function
Large Data Sets
Stochastic Processes
Model
Cover
Zero
Range of data
Class
Family
Nonseparability

Keywords

  • Convolution-based models
  • Non-separability
  • Spatio-temporal processes

ASJC Scopus subject areas

  • Statistics and Probability
  • Statistics, Probability and Uncertainty

Cite this

A Class of Convolution-Based Models for Spatio-Temporal Processes with Non-Separable Covariance Structure. / Rodrigues, Alexandre; Diggle, Peter J.

In: Scandinavian Journal of Statistics, Vol. 37, No. 4, 12.2010, p. 553-567.

Research output: Contribution to journalArticle

@article{f14d67c34c864c08920bbf9c86f12e78,
title = "A Class of Convolution-Based Models for Spatio-Temporal Processes with Non-Separable Covariance Structure",
abstract = "In this article, we propose a new parametric family of models for real-valued spatio-temporal stochastic processes S(x, t) and show how low-rank approximations can be used to overcome the computational problems that arise in fitting the proposed class of models to large datasets. Separable covariance models, in which the spatio-temporal covariance function of S(x, t) factorizes into a product of purely spatial and purely temporal functions, are often used as a convenient working assumption but are too inflexible to cover the range of covariance structures encountered in applications. We define positive and negative non-separability and show that in our proposed family we can capture positive, zero and negative non-separability by varying the value of a single parameter.",
keywords = "Convolution-based models, Non-separability, Spatio-temporal processes",
author = "Alexandre Rodrigues and Diggle, {Peter J.}",
year = "2010",
month = "12",
doi = "10.1111/j.1467-9469.2009.00675.x",
language = "English (US)",
volume = "37",
pages = "553--567",
journal = "Scandinavian Journal of Statistics",
issn = "0303-6898",
publisher = "Wiley-Blackwell",
number = "4",

}

TY - JOUR

T1 - A Class of Convolution-Based Models for Spatio-Temporal Processes with Non-Separable Covariance Structure

AU - Rodrigues, Alexandre

AU - Diggle, Peter J.

PY - 2010/12

Y1 - 2010/12

N2 - In this article, we propose a new parametric family of models for real-valued spatio-temporal stochastic processes S(x, t) and show how low-rank approximations can be used to overcome the computational problems that arise in fitting the proposed class of models to large datasets. Separable covariance models, in which the spatio-temporal covariance function of S(x, t) factorizes into a product of purely spatial and purely temporal functions, are often used as a convenient working assumption but are too inflexible to cover the range of covariance structures encountered in applications. We define positive and negative non-separability and show that in our proposed family we can capture positive, zero and negative non-separability by varying the value of a single parameter.

AB - In this article, we propose a new parametric family of models for real-valued spatio-temporal stochastic processes S(x, t) and show how low-rank approximations can be used to overcome the computational problems that arise in fitting the proposed class of models to large datasets. Separable covariance models, in which the spatio-temporal covariance function of S(x, t) factorizes into a product of purely spatial and purely temporal functions, are often used as a convenient working assumption but are too inflexible to cover the range of covariance structures encountered in applications. We define positive and negative non-separability and show that in our proposed family we can capture positive, zero and negative non-separability by varying the value of a single parameter.

KW - Convolution-based models

KW - Non-separability

KW - Spatio-temporal processes

UR - http://www.scopus.com/inward/record.url?scp=78349236578&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=78349236578&partnerID=8YFLogxK

U2 - 10.1111/j.1467-9469.2009.00675.x

DO - 10.1111/j.1467-9469.2009.00675.x

M3 - Article

VL - 37

SP - 553

EP - 567

JO - Scandinavian Journal of Statistics

JF - Scandinavian Journal of Statistics

SN - 0303-6898

IS - 4

ER -