Careful prior specification avoids incautious inference for log-Gaussian Cox point processes

Sigrunn H. S⊘rbye; Janine B. Illian; Daniel P. Simpson; David Burslem; Håvard Rue

doi:10.1111/rssc.12321

Careful prior specification avoids incautious inference for log-Gaussian Cox point processes

Sigrunn H. S⊘rbye^*, Janine B. Illian, Daniel P. Simpson, David Burslem, Håvard Rue

^*Corresponding author for this work

Statistics

Research output: Contribution to journal › Article › peer-review

16 Scopus citations

Abstract

Hyperprior specifications for random fields in spatial point process modelling can have a major influence on the results. In fitting log-Gaussian Cox processes to rainforest tree species, we consider a reparameterized model combining a spatially structured and an unstructured random field into a single component. This component has one hyperparameter accounting for marginal variance, whereas an additional hyperparameter governs the fraction of the variance that is explained by the spatially structured effect. This facilitates interpretation of the hyperparameters, and significance of covariates is studied for a range of hyperprior specifications. Appropriate scaling makes the analysis invariant to grid resolution.

Original language	English (US)
Pages (from-to)	543-564
Number of pages	22
Journal	Journal of the Royal Statistical Society. Series C: Applied Statistics
Volume	68
Issue number	3
DOIs	https://doi.org/10.1111/rssc.12321
State	Published - Apr 2019

Keywords

Bayesian analysis
Penalized complexity prior
R-INLA
Spatial modelling
Spatial point process

ASJC Scopus subject areas

Statistics and Probability
Statistics, Probability and Uncertainty

Access to Document

10.1111/rssc.12321

Cite this

@article{2cd9307037f3479ba10e0ec30f6cb409,

title = "Careful prior specification avoids incautious inference for log-Gaussian Cox point processes",

abstract = "Hyperprior specifications for random fields in spatial point process modelling can have a major influence on the results. In fitting log-Gaussian Cox processes to rainforest tree species, we consider a reparameterized model combining a spatially structured and an unstructured random field into a single component. This component has one hyperparameter accounting for marginal variance, whereas an additional hyperparameter governs the fraction of the variance that is explained by the spatially structured effect. This facilitates interpretation of the hyperparameters, and significance of covariates is studied for a range of hyperprior specifications. Appropriate scaling makes the analysis invariant to grid resolution.",

keywords = "Bayesian analysis, Penalized complexity prior, R-INLA, Spatial modelling, Spatial point process",

author = "S⊘rbye, {Sigrunn H.} and Illian, {Janine B.} and Simpson, {Daniel P.} and David Burslem and H{\aa}vard Rue",

note = "Publisher Copyright: {\textcopyright} 2018 Royal Statistical Society",

year = "2019",

month = apr,

doi = "10.1111/rssc.12321",

language = "English (US)",

volume = "68",

pages = "543--564",

journal = "Journal of the Royal Statistical Society. Series C: Applied Statistics",

issn = "0035-9254",

publisher = "Wiley-Blackwell",

number = "3",

}

TY - JOUR

T1 - Careful prior specification avoids incautious inference for log-Gaussian Cox point processes

AU - S⊘rbye, Sigrunn H.

AU - Illian, Janine B.

AU - Simpson, Daniel P.

AU - Burslem, David

AU - Rue, Håvard

PY - 2019/4

Y1 - 2019/4

N2 - Hyperprior specifications for random fields in spatial point process modelling can have a major influence on the results. In fitting log-Gaussian Cox processes to rainforest tree species, we consider a reparameterized model combining a spatially structured and an unstructured random field into a single component. This component has one hyperparameter accounting for marginal variance, whereas an additional hyperparameter governs the fraction of the variance that is explained by the spatially structured effect. This facilitates interpretation of the hyperparameters, and significance of covariates is studied for a range of hyperprior specifications. Appropriate scaling makes the analysis invariant to grid resolution.

AB - Hyperprior specifications for random fields in spatial point process modelling can have a major influence on the results. In fitting log-Gaussian Cox processes to rainforest tree species, we consider a reparameterized model combining a spatially structured and an unstructured random field into a single component. This component has one hyperparameter accounting for marginal variance, whereas an additional hyperparameter governs the fraction of the variance that is explained by the spatially structured effect. This facilitates interpretation of the hyperparameters, and significance of covariates is studied for a range of hyperprior specifications. Appropriate scaling makes the analysis invariant to grid resolution.

KW - Bayesian analysis

KW - Penalized complexity prior

KW - R-INLA

KW - Spatial modelling

KW - Spatial point process

UR - http://www.scopus.com/inward/record.url?scp=85055945954&partnerID=8YFLogxK

U2 - 10.1111/rssc.12321

DO - 10.1111/rssc.12321

M3 - Article

AN - SCOPUS:85055945954

SN - 0035-9254

VL - 68

SP - 543

EP - 564

JO - Journal of the Royal Statistical Society. Series C: Applied Statistics

JF - Journal of the Royal Statistical Society. Series C: Applied Statistics

IS - 3

ER -

Careful prior specification avoids incautious inference for log-Gaussian Cox point processes

Abstract

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this