Estimation and prediction for spatial generalized linear mixed models with parametric links via reparameterized importance sampling

Evangelos Evangelou, Vivekananda Roy

Research output: Contribution to journalArticle

Abstract

Spatial generalized linear mixed models (SGLMMs) are popular for analyzing non-Gaussian spatial data. These models assume a prescribed link function that relates the underlying spatial field with the mean response. There are circumstances, such as when the data contain outlying observations, where the use of a prescribed link function can result in poor fit, which can be improved by using a parametric link function. Some popular link functions, such as the Box-Cox, are unsuitable because they are inconsistent with the Gaussian assumption of the spatial field. We present sensible choices of parametric link functions which possess desirable properties. It is important to estimate the parameters of the link function, rather than assume a known value. To that end, we present a generalized importance sampling (GIS) estimator based on multiple Markov chains for empirical Bayes analysis of SGLMMs. The GIS estimator, although more efficient than the simple importance sampling, can be highly variable when used to estimate the parameters of certain link functions. Using suitable reparameterizations of the Monte Carlo samples, we propose modified GIS estimators that do not suffer from high variability. We use Laplace approximation for choosing the multiple importance densities in the GIS estimator. Finally, we develop a methodology for selecting models with appropriate link function family, which extends to choosing a spatial correlation function as well. We present an ensemble prediction of the mean response by appropriately weighting the estimates from different models. The proposed methodology is illustrated using simulated and real data examples.
LanguageEnglish
Pages289-315
Number of pages27
JournalSpatial Statistics
Volume29
Early online date18 Jan 2019
DOIs
StatusPublished - 1 Mar 2019

Keywords

  • Geostatistics
  • Laplace approximation
  • Markov chain Monte Carlo
  • Model selection
  • Multiple importance sampling
  • Reverse logistic regression

ASJC Scopus subject areas

  • Statistics and Probability
  • Computers in Earth Sciences
  • Management, Monitoring, Policy and Law

Cite this

@article{5988e5f0eaec44a88d28252d84b30c1c,
title = "Estimation and prediction for spatial generalized linear mixed models with parametric links via reparameterized importance sampling",
abstract = "Spatial generalized linear mixed models (SGLMMs) are popular for analyzing non-Gaussian spatial data. These models assume a prescribed link function that relates the underlying spatial field with the mean response. There are circumstances, such as when the data contain outlying observations, where the use of a prescribed link function can result in poor fit, which can be improved by using a parametric link function. Some popular link functions, such as the Box-Cox, are unsuitable because they are inconsistent with the Gaussian assumption of the spatial field. We present sensible choices of parametric link functions which possess desirable properties. It is important to estimate the parameters of the link function, rather than assume a known value. To that end, we present a generalized importance sampling (GIS) estimator based on multiple Markov chains for empirical Bayes analysis of SGLMMs. The GIS estimator, although more efficient than the simple importance sampling, can be highly variable when used to estimate the parameters of certain link functions. Using suitable reparameterizations of the Monte Carlo samples, we propose modified GIS estimators that do not suffer from high variability. We use Laplace approximation for choosing the multiple importance densities in the GIS estimator. Finally, we develop a methodology for selecting models with appropriate link function family, which extends to choosing a spatial correlation function as well. We present an ensemble prediction of the mean response by appropriately weighting the estimates from different models. The proposed methodology is illustrated using simulated and real data examples.",
keywords = "Geostatistics, Laplace approximation, Markov chain Monte Carlo, Model selection, Multiple importance sampling, Reverse logistic regression",
author = "Evangelos Evangelou and Vivekananda Roy",
year = "2019",
month = "3",
day = "1",
doi = "10.1016/j.spasta.2019.01.005",
language = "English",
volume = "29",
pages = "289--315",
journal = "Spatial Statistics",
issn = "2211-6753",
publisher = "Elsevier",

}

TY - JOUR

T1 - Estimation and prediction for spatial generalized linear mixed models with parametric links via reparameterized importance sampling

AU - Evangelou, Evangelos

AU - Roy, Vivekananda

PY - 2019/3/1

Y1 - 2019/3/1

N2 - Spatial generalized linear mixed models (SGLMMs) are popular for analyzing non-Gaussian spatial data. These models assume a prescribed link function that relates the underlying spatial field with the mean response. There are circumstances, such as when the data contain outlying observations, where the use of a prescribed link function can result in poor fit, which can be improved by using a parametric link function. Some popular link functions, such as the Box-Cox, are unsuitable because they are inconsistent with the Gaussian assumption of the spatial field. We present sensible choices of parametric link functions which possess desirable properties. It is important to estimate the parameters of the link function, rather than assume a known value. To that end, we present a generalized importance sampling (GIS) estimator based on multiple Markov chains for empirical Bayes analysis of SGLMMs. The GIS estimator, although more efficient than the simple importance sampling, can be highly variable when used to estimate the parameters of certain link functions. Using suitable reparameterizations of the Monte Carlo samples, we propose modified GIS estimators that do not suffer from high variability. We use Laplace approximation for choosing the multiple importance densities in the GIS estimator. Finally, we develop a methodology for selecting models with appropriate link function family, which extends to choosing a spatial correlation function as well. We present an ensemble prediction of the mean response by appropriately weighting the estimates from different models. The proposed methodology is illustrated using simulated and real data examples.

AB - Spatial generalized linear mixed models (SGLMMs) are popular for analyzing non-Gaussian spatial data. These models assume a prescribed link function that relates the underlying spatial field with the mean response. There are circumstances, such as when the data contain outlying observations, where the use of a prescribed link function can result in poor fit, which can be improved by using a parametric link function. Some popular link functions, such as the Box-Cox, are unsuitable because they are inconsistent with the Gaussian assumption of the spatial field. We present sensible choices of parametric link functions which possess desirable properties. It is important to estimate the parameters of the link function, rather than assume a known value. To that end, we present a generalized importance sampling (GIS) estimator based on multiple Markov chains for empirical Bayes analysis of SGLMMs. The GIS estimator, although more efficient than the simple importance sampling, can be highly variable when used to estimate the parameters of certain link functions. Using suitable reparameterizations of the Monte Carlo samples, we propose modified GIS estimators that do not suffer from high variability. We use Laplace approximation for choosing the multiple importance densities in the GIS estimator. Finally, we develop a methodology for selecting models with appropriate link function family, which extends to choosing a spatial correlation function as well. We present an ensemble prediction of the mean response by appropriately weighting the estimates from different models. The proposed methodology is illustrated using simulated and real data examples.

KW - Geostatistics

KW - Laplace approximation

KW - Markov chain Monte Carlo

KW - Model selection

KW - Multiple importance sampling

KW - Reverse logistic regression

UR - http://www.scopus.com/inward/record.url?scp=85060469122&partnerID=8YFLogxK

U2 - 10.1016/j.spasta.2019.01.005

DO - 10.1016/j.spasta.2019.01.005

M3 - Article

VL - 29

SP - 289

EP - 315

JO - Spatial Statistics

T2 - Spatial Statistics

JF - Spatial Statistics

SN - 2211-6753

ER -