No Cover Image

Journal article 537 views 287 downloads

Dynamic generalised additive models (DGAMs) for forecasting discrete ecological time series

Nicholas J. Clark Orcid Logo, Konstans Wells Orcid Logo

Methods in Ecology and Evolution, Volume: 14, Issue: 3

Swansea University Author: Konstans Wells Orcid Logo

  • Clark_etal_2022_MethodsEcolEvol.pdf

    PDF | Version of Record

    © 2022 The Authors. This is an open access article under the terms of the Creative Commons Attribution License.

    Download (1.61MB)

Check full text

DOI (Published version): 10.1111/2041-210x.13974

Abstract

1. Generalized Additive Models (GAMs) are increasingly popular tools for estimating smooth nonlinear relationships between predictors and response variables. GAMs are particularly relevant in ecology for representing hierarchical functions for discrete responses that encompass complex features inclu...

Full description

Published in: Methods in Ecology and Evolution
ISSN: 2041-210X 2041-210X
Published: Wiley Wiley 2022
Online Access: Check full text

URI: https://cronfa.swan.ac.uk/Record/cronfa60707
first_indexed 2022-08-03T21:35:12Z
last_indexed 2024-11-14T12:17:49Z
id cronfa60707
recordtype SURis
fullrecord <?xml version="1.0"?><rfc1807><datestamp>2023-06-12T16:13:52.9671420</datestamp><bib-version>v2</bib-version><id>60707</id><entry>2022-08-02</entry><title>Dynamic generalised additive models (DGAMs) for forecasting discrete ecological time series</title><swanseaauthors><author><sid>d18166c31e89833c55ef0f2cbb551243</sid><ORCID>0000-0003-0377-2463</ORCID><firstname>Konstans</firstname><surname>Wells</surname><name>Konstans Wells</name><active>true</active><ethesisStudent>false</ethesisStudent></author></swanseaauthors><date>2022-08-02</date><deptcode>BGPS</deptcode><abstract>1. Generalized Additive Models (GAMs) are increasingly popular tools for estimating smooth nonlinear relationships between predictors and response variables. GAMs are particularly relevant in ecology for representing hierarchical functions for discrete responses that encompass complex features including zero-inflation, truncation and uneven sampling. However, GAMs are less useful for producing forecasts as their smooth functions provide unstable predictions outside the range of training data.2. We introduce Dynamic Generalized Additive Models (DGAMs), where the GAM linear predictor is jointly estimated with unobserved dynamic components to model time series that evolve as a function of nonlinear predictor associations and latent temporal processes. These models are especially useful for analysing multiple series, as they can estimate hierarchical smooth functions while learning complex temporal associations via dimension-reduced latent factor processes. We implement our models in the mvgam R package, which estimates unobserved parameters for smoothing splines and latent temporal processes in a probabilistic framework.3. Using simulations, we illustrate how our models outperform competing formulations in realistic ecological forecasting tasks while identifying important smooth predictor functions. We use a real-world case study to highlight some of mvgam&#x2019;s key features, which include functions for: calculating correlations among series&#x2019; latent trends, performing model selection using rolling window forecasts and posterior predictive checks, online data augmentation via a recursive particle filter, and visualising probabilistic uncertainties for smooth functions and predictions.4. Dynamic GAMs (DGAMs) offer a solution to the challenge of forecasting discrete time series while estimating ecologically relevant nonlinear predictor associations. Our Bayesian latent factor approach will be particularly useful for exploring competing dynamic ecological models that encompass hierarchical smoothing structures while providing forecasts with robust uncertainties, tasks that are becoming increasingly important in applied ecology</abstract><type>Journal Article</type><journal>Methods in Ecology and Evolution</journal><volume>14</volume><journalNumber>3</journalNumber><paginationStart/><paginationEnd/><publisher>Wiley</publisher><placeOfPublication>Wiley</placeOfPublication><isbnPrint/><isbnElectronic/><issnPrint>2041-210X</issnPrint><issnElectronic>2041-210X</issnElectronic><keywords>dynamic factor model, ecological forecasting, generalised additive model, hierarchical model, JAGS, R package, Stan</keywords><publishedDay>11</publishedDay><publishedMonth>9</publishedMonth><publishedYear>2022</publishedYear><publishedDate>2022-09-11</publishedDate><doi>10.1111/2041-210x.13974</doi><url/><notes/><college>COLLEGE NANME</college><department>Biosciences Geography and Physics School</department><CollegeCode>COLLEGE CODE</CollegeCode><DepartmentCode>BGPS</DepartmentCode><institution>Swansea University</institution><apcterm/><funders>Australian Research Council. Grant Number: DE210101439</funders><projectreference/><lastEdited>2023-06-12T16:13:52.9671420</lastEdited><Created>2022-08-02T15:22:06.9299044</Created><path><level id="1">Faculty of Science and Engineering</level><level id="2">School of Biosciences, Geography and Physics - Biosciences</level></path><authors><author><firstname>Nicholas J.</firstname><surname>Clark</surname><orcid>0000-0001-7131-3301</orcid><order>1</order></author><author><firstname>Konstans</firstname><surname>Wells</surname><orcid>0000-0003-0377-2463</orcid><order>2</order></author></authors><documents><document><filename>60707__25125__649684b094ca47baa69d24156bfe8e79.pdf</filename><originalFilename>Clark_etal_2022_MethodsEcolEvol.pdf</originalFilename><uploaded>2022-09-12T17:25:46.3471729</uploaded><type>Output</type><contentLength>1686689</contentLength><contentType>application/pdf</contentType><version>Version of Record</version><cronfaStatus>true</cronfaStatus><documentNotes>&#xA9; 2022 The Authors. This is an open access article under the terms of the Creative Commons Attribution License.</documentNotes><copyrightCorrect>true</copyrightCorrect><language>eng</language><licence>http://creativecommons.org/licenses/by/4.0/</licence></document></documents><OutputDurs/></rfc1807>
spelling 2023-06-12T16:13:52.9671420 v2 60707 2022-08-02 Dynamic generalised additive models (DGAMs) for forecasting discrete ecological time series d18166c31e89833c55ef0f2cbb551243 0000-0003-0377-2463 Konstans Wells Konstans Wells true false 2022-08-02 BGPS 1. Generalized Additive Models (GAMs) are increasingly popular tools for estimating smooth nonlinear relationships between predictors and response variables. GAMs are particularly relevant in ecology for representing hierarchical functions for discrete responses that encompass complex features including zero-inflation, truncation and uneven sampling. However, GAMs are less useful for producing forecasts as their smooth functions provide unstable predictions outside the range of training data.2. We introduce Dynamic Generalized Additive Models (DGAMs), where the GAM linear predictor is jointly estimated with unobserved dynamic components to model time series that evolve as a function of nonlinear predictor associations and latent temporal processes. These models are especially useful for analysing multiple series, as they can estimate hierarchical smooth functions while learning complex temporal associations via dimension-reduced latent factor processes. We implement our models in the mvgam R package, which estimates unobserved parameters for smoothing splines and latent temporal processes in a probabilistic framework.3. Using simulations, we illustrate how our models outperform competing formulations in realistic ecological forecasting tasks while identifying important smooth predictor functions. We use a real-world case study to highlight some of mvgam’s key features, which include functions for: calculating correlations among series’ latent trends, performing model selection using rolling window forecasts and posterior predictive checks, online data augmentation via a recursive particle filter, and visualising probabilistic uncertainties for smooth functions and predictions.4. Dynamic GAMs (DGAMs) offer a solution to the challenge of forecasting discrete time series while estimating ecologically relevant nonlinear predictor associations. Our Bayesian latent factor approach will be particularly useful for exploring competing dynamic ecological models that encompass hierarchical smoothing structures while providing forecasts with robust uncertainties, tasks that are becoming increasingly important in applied ecology Journal Article Methods in Ecology and Evolution 14 3 Wiley Wiley 2041-210X 2041-210X dynamic factor model, ecological forecasting, generalised additive model, hierarchical model, JAGS, R package, Stan 11 9 2022 2022-09-11 10.1111/2041-210x.13974 COLLEGE NANME Biosciences Geography and Physics School COLLEGE CODE BGPS Swansea University Australian Research Council. Grant Number: DE210101439 2023-06-12T16:13:52.9671420 2022-08-02T15:22:06.9299044 Faculty of Science and Engineering School of Biosciences, Geography and Physics - Biosciences Nicholas J. Clark 0000-0001-7131-3301 1 Konstans Wells 0000-0003-0377-2463 2 60707__25125__649684b094ca47baa69d24156bfe8e79.pdf Clark_etal_2022_MethodsEcolEvol.pdf 2022-09-12T17:25:46.3471729 Output 1686689 application/pdf Version of Record true © 2022 The Authors. This is an open access article under the terms of the Creative Commons Attribution License. true eng http://creativecommons.org/licenses/by/4.0/
title Dynamic generalised additive models (DGAMs) for forecasting discrete ecological time series
spellingShingle Dynamic generalised additive models (DGAMs) for forecasting discrete ecological time series
Konstans Wells
title_short Dynamic generalised additive models (DGAMs) for forecasting discrete ecological time series
title_full Dynamic generalised additive models (DGAMs) for forecasting discrete ecological time series
title_fullStr Dynamic generalised additive models (DGAMs) for forecasting discrete ecological time series
title_full_unstemmed Dynamic generalised additive models (DGAMs) for forecasting discrete ecological time series
title_sort Dynamic generalised additive models (DGAMs) for forecasting discrete ecological time series
author_id_str_mv d18166c31e89833c55ef0f2cbb551243
author_id_fullname_str_mv d18166c31e89833c55ef0f2cbb551243_***_Konstans Wells
author Konstans Wells
author2 Nicholas J. Clark
Konstans Wells
format Journal article
container_title Methods in Ecology and Evolution
container_volume 14
container_issue 3
publishDate 2022
institution Swansea University
issn 2041-210X
2041-210X
doi_str_mv 10.1111/2041-210x.13974
publisher Wiley
college_str Faculty of Science and Engineering
hierarchytype
hierarchy_top_id facultyofscienceandengineering
hierarchy_top_title Faculty of Science and Engineering
hierarchy_parent_id facultyofscienceandengineering
hierarchy_parent_title Faculty of Science and Engineering
department_str School of Biosciences, Geography and Physics - Biosciences{{{_:::_}}}Faculty of Science and Engineering{{{_:::_}}}School of Biosciences, Geography and Physics - Biosciences
document_store_str 1
active_str 0
description 1. Generalized Additive Models (GAMs) are increasingly popular tools for estimating smooth nonlinear relationships between predictors and response variables. GAMs are particularly relevant in ecology for representing hierarchical functions for discrete responses that encompass complex features including zero-inflation, truncation and uneven sampling. However, GAMs are less useful for producing forecasts as their smooth functions provide unstable predictions outside the range of training data.2. We introduce Dynamic Generalized Additive Models (DGAMs), where the GAM linear predictor is jointly estimated with unobserved dynamic components to model time series that evolve as a function of nonlinear predictor associations and latent temporal processes. These models are especially useful for analysing multiple series, as they can estimate hierarchical smooth functions while learning complex temporal associations via dimension-reduced latent factor processes. We implement our models in the mvgam R package, which estimates unobserved parameters for smoothing splines and latent temporal processes in a probabilistic framework.3. Using simulations, we illustrate how our models outperform competing formulations in realistic ecological forecasting tasks while identifying important smooth predictor functions. We use a real-world case study to highlight some of mvgam’s key features, which include functions for: calculating correlations among series’ latent trends, performing model selection using rolling window forecasts and posterior predictive checks, online data augmentation via a recursive particle filter, and visualising probabilistic uncertainties for smooth functions and predictions.4. Dynamic GAMs (DGAMs) offer a solution to the challenge of forecasting discrete time series while estimating ecologically relevant nonlinear predictor associations. Our Bayesian latent factor approach will be particularly useful for exploring competing dynamic ecological models that encompass hierarchical smoothing structures while providing forecasts with robust uncertainties, tasks that are becoming increasingly important in applied ecology
published_date 2022-09-11T05:17:43Z
_version_ 1821381401866403840
score 11.3749895

Similar Items