Variable selection using penalised likelihoods for point patterns on a linear network.

Motivated by the analysis of a comprehensive database of road traffic accidents, we investigate methods of variable selection for spatial point process models on a linear network. The original data may include explanatory spatial covariates, such as road curvature, and ‘mark’ variables attributed to...

Full description

Bibliographic Details
Main Authors: Rakshit, Suman, McSwiggan, Greg, Nair, Gopalan, Baddeley, Adrian
Format: Journal Article
Language:English
Published: Wiley-Blackwell 2021
Subjects:
Online Access:http://purl.org/au-research/grants/arc/DP130102322
http://hdl.handle.net/20.500.11937/91582
_version_ 1848765554751438848
author Rakshit, Suman
McSwiggan, Greg
Nair, Gopalan
Baddeley, Adrian
author_facet Rakshit, Suman
McSwiggan, Greg
Nair, Gopalan
Baddeley, Adrian
author_sort Rakshit, Suman
building Curtin Institutional Repository
collection Online Access
description Motivated by the analysis of a comprehensive database of road traffic accidents, we investigate methods of variable selection for spatial point process models on a linear network. The original data may include explanatory spatial covariates, such as road curvature, and ‘mark’ variables attributed to individual accidents, such as accident severity. The treatment of mark variables is new. Variable selection is applied to the canonical covariates, which may include spatial covariate effects, mark effects and mark-covariate interactions. We approximate the likelihood of the point process model by that of a generalised linear model, in such a way that spatial covariates and marks are both associated with canonical covariates. We impose a convex penalty on the log likelihood, principally the elastic-net penalty, and maximise the penalised loglikelihood by cyclic coordinate ascent. A simulation study compares the performances of the lasso, ridge regression and elastic-net methods of variable selection on their ability to select variables correctly, and on their bias and standard error. Standard techniques for selecting the regularisation parameter γ often yielded unsatisfactory results. We propose two new rules for selecting γ which are designed to have better performance. The methods are tested on a small dataset on crimes in a Chicago neighbourhood, and applied to a large dataset of road traffic accidents in Western Australia.
first_indexed 2025-11-14T11:37:06Z
format Journal Article
id curtin-20.500.11937-91582
institution Curtin University Malaysia
institution_category Local University
language English
last_indexed 2025-11-14T11:37:06Z
publishDate 2021
publisher Wiley-Blackwell
recordtype eprints
repository_type Digital Repository
spelling curtin-20.500.11937-915822023-05-17T04:44:33Z Variable selection using penalised likelihoods for point patterns on a linear network. Rakshit, Suman McSwiggan, Greg Nair, Gopalan Baddeley, Adrian Science & Technology Physical Sciences Statistics & Probability Mathematics discretised models elastic net generalised linear model lasso Poisson process SPATIAL STATISTICAL-MODELS 2ND-ORDER ANALYSIS REGULARIZATION EQUIVALENCE REGRESSION GRAPHS LASSO Motivated by the analysis of a comprehensive database of road traffic accidents, we investigate methods of variable selection for spatial point process models on a linear network. The original data may include explanatory spatial covariates, such as road curvature, and ‘mark’ variables attributed to individual accidents, such as accident severity. The treatment of mark variables is new. Variable selection is applied to the canonical covariates, which may include spatial covariate effects, mark effects and mark-covariate interactions. We approximate the likelihood of the point process model by that of a generalised linear model, in such a way that spatial covariates and marks are both associated with canonical covariates. We impose a convex penalty on the log likelihood, principally the elastic-net penalty, and maximise the penalised loglikelihood by cyclic coordinate ascent. A simulation study compares the performances of the lasso, ridge regression and elastic-net methods of variable selection on their ability to select variables correctly, and on their bias and standard error. Standard techniques for selecting the regularisation parameter γ often yielded unsatisfactory results. We propose two new rules for selecting γ which are designed to have better performance. The methods are tested on a small dataset on crimes in a Chicago neighbourhood, and applied to a large dataset of road traffic accidents in Western Australia. 2021 Journal Article http://hdl.handle.net/20.500.11937/91582 10.1111/anzs.12341 English http://purl.org/au-research/grants/arc/DP130102322 Wiley-Blackwell fulltext
spellingShingle Science & Technology
Physical Sciences
Statistics & Probability
Mathematics
discretised models
elastic net
generalised linear model
lasso
Poisson process
SPATIAL STATISTICAL-MODELS
2ND-ORDER ANALYSIS
REGULARIZATION
EQUIVALENCE
REGRESSION
GRAPHS
LASSO
Rakshit, Suman
McSwiggan, Greg
Nair, Gopalan
Baddeley, Adrian
Variable selection using penalised likelihoods for point patterns on a linear network.
title Variable selection using penalised likelihoods for point patterns on a linear network.
title_full Variable selection using penalised likelihoods for point patterns on a linear network.
title_fullStr Variable selection using penalised likelihoods for point patterns on a linear network.
title_full_unstemmed Variable selection using penalised likelihoods for point patterns on a linear network.
title_short Variable selection using penalised likelihoods for point patterns on a linear network.
title_sort variable selection using penalised likelihoods for point patterns on a linear network.
topic Science & Technology
Physical Sciences
Statistics & Probability
Mathematics
discretised models
elastic net
generalised linear model
lasso
Poisson process
SPATIAL STATISTICAL-MODELS
2ND-ORDER ANALYSIS
REGULARIZATION
EQUIVALENCE
REGRESSION
GRAPHS
LASSO
url http://purl.org/au-research/grants/arc/DP130102322
http://hdl.handle.net/20.500.11937/91582