Dynamic Bayesian models: extensions and new proposals
Autor(a) principal: | |
---|---|
Data de Publicação: | 2019 |
Tipo de documento: | Tese |
Idioma: | eng |
Título da fonte: | Repositório Institucional da UFMG |
Texto Completo: | http://hdl.handle.net/1843/33546 |
Resumo: | Even though many time series presents problems such as overdispersion, zero inflation and change-points, these features, usually, are not incorporated into the most common dynamic Bayesian models available in the literature. To address these problems, we worked on two strands in this dissertation. In the first strand, the objective is to introduce new Bayesian dynamic models for time series of counts that allow for observations in distributions that can more adequately adjust to some common features related to the modeling of discrete data. We present a new framework for uniparametric Dynamic Bayesian Models of counts whose particular cases include Bell, Poisson-Lindley, Yule-Simon and Borel models. Furthermore, a biparametric Negative binomial model with unknown shape parameter is provided. The inferential procedure preserves the sequential nature of the Bayesian analysis and is similar to the Dynamic Generalized Linear Models (DGLM) with a novel of incorporating Monte Carlo integration to the recursive algorithm in order to deal with the intractability of the updating distributions and an ARMS step to sample from the posterior distribution of the shape parameter. We also consider a conjugate Beta Prime of the second kind distribution prior for the mean of the process. The simulation results show a good performance of the estimators considered for the static parameter, which can be reasonably estimated. The application results also highlights a better performance of the proposed uni/biparametric models over the Poisson model. In the second strand of this work we incorporate the Product Partition Models class into the DGLM. This new formulation, that we call DGLM-PPM, retains the flexibility and generality of the DGLM class and also provides a framework for Bayesian multiple change-point detection in time series. To sample from the partition and the discount factor we use a Gibbs Sampler with an ARMS step appended. A simulation study is conducted and the results show that the proposed model is able to detect the points of regime switch in the simulated data. The superiority of our proposal over the conventional DGLM is further confirmed in two real data applications in which the DGLM-PPM outperforms the conventional DGLM in-sample and out-of-sample. |
id |
UFMG_fc935c9b75d4af314db7a721e8749ba5 |
---|---|
oai_identifier_str |
oai:repositorio.ufmg.br:1843/33546 |
network_acronym_str |
UFMG |
network_name_str |
Repositório Institucional da UFMG |
repository_id_str |
|
spelling |
Thiago Rezende dos Santoshttp://lattes.cnpq.br/9458275921031976Fábio Nogueira DemarquiGlaura da Conceição FrancoRosângela Helena LoschiDani GamermanAlessandro Queiroz José Samagliahttp://lattes.cnpq.br/1856533424491195Victor Schmidt Comitti2020-05-26T19:11:23Z2020-05-26T19:11:23Z2019-07-05http://hdl.handle.net/1843/33546Even though many time series presents problems such as overdispersion, zero inflation and change-points, these features, usually, are not incorporated into the most common dynamic Bayesian models available in the literature. To address these problems, we worked on two strands in this dissertation. In the first strand, the objective is to introduce new Bayesian dynamic models for time series of counts that allow for observations in distributions that can more adequately adjust to some common features related to the modeling of discrete data. We present a new framework for uniparametric Dynamic Bayesian Models of counts whose particular cases include Bell, Poisson-Lindley, Yule-Simon and Borel models. Furthermore, a biparametric Negative binomial model with unknown shape parameter is provided. The inferential procedure preserves the sequential nature of the Bayesian analysis and is similar to the Dynamic Generalized Linear Models (DGLM) with a novel of incorporating Monte Carlo integration to the recursive algorithm in order to deal with the intractability of the updating distributions and an ARMS step to sample from the posterior distribution of the shape parameter. We also consider a conjugate Beta Prime of the second kind distribution prior for the mean of the process. The simulation results show a good performance of the estimators considered for the static parameter, which can be reasonably estimated. The application results also highlights a better performance of the proposed uni/biparametric models over the Poisson model. In the second strand of this work we incorporate the Product Partition Models class into the DGLM. This new formulation, that we call DGLM-PPM, retains the flexibility and generality of the DGLM class and also provides a framework for Bayesian multiple change-point detection in time series. To sample from the partition and the discount factor we use a Gibbs Sampler with an ARMS step appended. A simulation study is conducted and the results show that the proposed model is able to detect the points of regime switch in the simulated data. The superiority of our proposal over the conventional DGLM is further confirmed in two real data applications in which the DGLM-PPM outperforms the conventional DGLM in-sample and out-of-sample.Embora muitas séries temporais apresentem problemas como superdispersão, inflação zero e pontos de mudança, essas características, geralmente, não são incorporadas aos modelos Bayesianos dinâmicos mais comuns disponíveis na literatura. Para resolver esses problemas, trabalhamos em duas vertentes nesta tese. Na primeira vertente, o objetivo é introduzir novos modelos dinâmicos Bayesianos para séries temporais de contagem que permitam observações em distribuições que se ajustam melhor a algumas características comuns relacionadas à modelagem de dados discretos. Apresentamos uma nova estrutura para modelos dinâmicos Bayesianos uniparamétricos de contagem cujos casos particulares incluem os modelos Bell, Poisson-Lindley, Yule-Simon e Borel. Além disso, propomos um modelo binomial negativo biparamétrico com parâmetro de forma desconhecido. O procedimento de inferência preserva a natureza seqüencial da análise Bayesiana e é semelhante ao dos Modelos Lineares Generalizados Dinâmicos (DGLM). Nossa proposta incorpora passos de integração Monte Carlo ao algoritmo recursivo para lidar com a intratabilidade das distribuições de atualização e um passo de ARMS para amostrar da distribuição a posteriori do parâmetro de forma. Também consideramos uma distribuição conjugada Beta Prime do segundo tipo para a média do processo. Os resultados de simulação mostram um bom desempenho dos estimadores considerados para o parâmetro estático do modelo mostrando que ele pode ser razoavelmente estimado. Os resultados da aplicação também destacam um melhor desempenho dos modelos uni / biparamétricos propostos sobre o modelo Poisson. Na segunda vertente deste trabalho, incorporamos a classe de Modelos de Partição Produto ao DGLM. Essa nova formulação, aqui chamada de DGLM-PPM, retém a flexibilidade e a generalidade da classe DGLM e também fornece uma estrutura para detecção de múltiplos pontos de mudança em séries temporais. Um estudo de simulação é realizado e os resultados mostram que o modelo proposto é capaz de detectar os pontos de mudança de regime nos dados simulados. A superioridade de nossa proposta em relação ao DGLM convencional é confirmada em duas aplicações a dados reais nas quais o DGLM-PPM supera o DGLM convencional em performance dentro e fora da amostra.CAPES - Coordenação de Aperfeiçoamento de Pessoal de Nível SuperiorengUniversidade Federal de Minas GeraisPrograma de Pós-Graduação em EstatísticaUFMGBrasilICX - DEPARTAMENTO DE ESTATÍSTICAhttp://creativecommons.org/licenses/by-nc-nd/3.0/pt/info:eu-repo/semantics/openAccessEstatística. - Teses.Séries temporais.Modelos dinâmicos Bayesianos.Modelos de partição produto.Dados de contagemzero inflated/overdispersed distributionNegative Binomial distributionParameter driven modelDynamic Generalized Linear ModelBayesian inferencechange-point detectionProduct Partition ModelsDynamic Bayesian models: extensions and new proposalsModelos dinâmicos Bayesianos: extensões e novas propostasinfo:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/doctoralThesisreponame:Repositório Institucional da UFMGinstname:Universidade Federal de Minas Gerais (UFMG)instacron:UFMGORIGINALTESE_vers_o_final (1).pdfTESE_vers_o_final (1).pdfapplication/pdf2016243https://repositorio.ufmg.br/bitstream/1843/33546/1/TESE_vers_o_final%20%281%29.pdfe768061ee5aa432d1820807855a2fe60MD51CC-LICENSElicense_rdflicense_rdfapplication/rdf+xml; charset=utf-8811https://repositorio.ufmg.br/bitstream/1843/33546/2/license_rdfcfd6801dba008cb6adbd9838b81582abMD52LICENSElicense.txtlicense.txttext/plain; charset=utf-82119https://repositorio.ufmg.br/bitstream/1843/33546/3/license.txt34badce4be7e31e3adb4575ae96af679MD531843/335462020-05-26 16:11:23.084oai:repositorio.ufmg.br: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Repositório de PublicaçõesPUBhttps://repositorio.ufmg.br/oaiopendoar:2020-05-26T19:11:23Repositório Institucional da UFMG - Universidade Federal de Minas Gerais (UFMG)false |
dc.title.pt_BR.fl_str_mv |
Dynamic Bayesian models: extensions and new proposals |
dc.title.alternative.pt_BR.fl_str_mv |
Modelos dinâmicos Bayesianos: extensões e novas propostas |
title |
Dynamic Bayesian models: extensions and new proposals |
spellingShingle |
Dynamic Bayesian models: extensions and new proposals Victor Schmidt Comitti zero inflated/overdispersed distribution Negative Binomial distribution Parameter driven model Dynamic Generalized Linear Model Bayesian inference change-point detection Product Partition Models Estatística. - Teses. Séries temporais. Modelos dinâmicos Bayesianos. Modelos de partição produto. Dados de contagem |
title_short |
Dynamic Bayesian models: extensions and new proposals |
title_full |
Dynamic Bayesian models: extensions and new proposals |
title_fullStr |
Dynamic Bayesian models: extensions and new proposals |
title_full_unstemmed |
Dynamic Bayesian models: extensions and new proposals |
title_sort |
Dynamic Bayesian models: extensions and new proposals |
author |
Victor Schmidt Comitti |
author_facet |
Victor Schmidt Comitti |
author_role |
author |
dc.contributor.advisor1.fl_str_mv |
Thiago Rezende dos Santos |
dc.contributor.advisor1Lattes.fl_str_mv |
http://lattes.cnpq.br/9458275921031976 |
dc.contributor.advisor-co1.fl_str_mv |
Fábio Nogueira Demarqui |
dc.contributor.referee1.fl_str_mv |
Glaura da Conceição Franco |
dc.contributor.referee2.fl_str_mv |
Rosângela Helena Loschi |
dc.contributor.referee3.fl_str_mv |
Dani Gamerman |
dc.contributor.referee4.fl_str_mv |
Alessandro Queiroz José Samaglia |
dc.contributor.authorLattes.fl_str_mv |
http://lattes.cnpq.br/1856533424491195 |
dc.contributor.author.fl_str_mv |
Victor Schmidt Comitti |
contributor_str_mv |
Thiago Rezende dos Santos Fábio Nogueira Demarqui Glaura da Conceição Franco Rosângela Helena Loschi Dani Gamerman Alessandro Queiroz José Samaglia |
dc.subject.por.fl_str_mv |
zero inflated/overdispersed distribution Negative Binomial distribution Parameter driven model Dynamic Generalized Linear Model Bayesian inference change-point detection Product Partition Models |
topic |
zero inflated/overdispersed distribution Negative Binomial distribution Parameter driven model Dynamic Generalized Linear Model Bayesian inference change-point detection Product Partition Models Estatística. - Teses. Séries temporais. Modelos dinâmicos Bayesianos. Modelos de partição produto. Dados de contagem |
dc.subject.other.pt_BR.fl_str_mv |
Estatística. - Teses. Séries temporais. Modelos dinâmicos Bayesianos. Modelos de partição produto. Dados de contagem |
description |
Even though many time series presents problems such as overdispersion, zero inflation and change-points, these features, usually, are not incorporated into the most common dynamic Bayesian models available in the literature. To address these problems, we worked on two strands in this dissertation. In the first strand, the objective is to introduce new Bayesian dynamic models for time series of counts that allow for observations in distributions that can more adequately adjust to some common features related to the modeling of discrete data. We present a new framework for uniparametric Dynamic Bayesian Models of counts whose particular cases include Bell, Poisson-Lindley, Yule-Simon and Borel models. Furthermore, a biparametric Negative binomial model with unknown shape parameter is provided. The inferential procedure preserves the sequential nature of the Bayesian analysis and is similar to the Dynamic Generalized Linear Models (DGLM) with a novel of incorporating Monte Carlo integration to the recursive algorithm in order to deal with the intractability of the updating distributions and an ARMS step to sample from the posterior distribution of the shape parameter. We also consider a conjugate Beta Prime of the second kind distribution prior for the mean of the process. The simulation results show a good performance of the estimators considered for the static parameter, which can be reasonably estimated. The application results also highlights a better performance of the proposed uni/biparametric models over the Poisson model. In the second strand of this work we incorporate the Product Partition Models class into the DGLM. This new formulation, that we call DGLM-PPM, retains the flexibility and generality of the DGLM class and also provides a framework for Bayesian multiple change-point detection in time series. To sample from the partition and the discount factor we use a Gibbs Sampler with an ARMS step appended. A simulation study is conducted and the results show that the proposed model is able to detect the points of regime switch in the simulated data. The superiority of our proposal over the conventional DGLM is further confirmed in two real data applications in which the DGLM-PPM outperforms the conventional DGLM in-sample and out-of-sample. |
publishDate |
2019 |
dc.date.issued.fl_str_mv |
2019-07-05 |
dc.date.accessioned.fl_str_mv |
2020-05-26T19:11:23Z |
dc.date.available.fl_str_mv |
2020-05-26T19:11:23Z |
dc.type.status.fl_str_mv |
info:eu-repo/semantics/publishedVersion |
dc.type.driver.fl_str_mv |
info:eu-repo/semantics/doctoralThesis |
format |
doctoralThesis |
status_str |
publishedVersion |
dc.identifier.uri.fl_str_mv |
http://hdl.handle.net/1843/33546 |
url |
http://hdl.handle.net/1843/33546 |
dc.language.iso.fl_str_mv |
eng |
language |
eng |
dc.rights.driver.fl_str_mv |
http://creativecommons.org/licenses/by-nc-nd/3.0/pt/ info:eu-repo/semantics/openAccess |
rights_invalid_str_mv |
http://creativecommons.org/licenses/by-nc-nd/3.0/pt/ |
eu_rights_str_mv |
openAccess |
dc.publisher.none.fl_str_mv |
Universidade Federal de Minas Gerais |
dc.publisher.program.fl_str_mv |
Programa de Pós-Graduação em Estatística |
dc.publisher.initials.fl_str_mv |
UFMG |
dc.publisher.country.fl_str_mv |
Brasil |
dc.publisher.department.fl_str_mv |
ICX - DEPARTAMENTO DE ESTATÍSTICA |
publisher.none.fl_str_mv |
Universidade Federal de Minas Gerais |
dc.source.none.fl_str_mv |
reponame:Repositório Institucional da UFMG instname:Universidade Federal de Minas Gerais (UFMG) instacron:UFMG |
instname_str |
Universidade Federal de Minas Gerais (UFMG) |
instacron_str |
UFMG |
institution |
UFMG |
reponame_str |
Repositório Institucional da UFMG |
collection |
Repositório Institucional da UFMG |
bitstream.url.fl_str_mv |
https://repositorio.ufmg.br/bitstream/1843/33546/1/TESE_vers_o_final%20%281%29.pdf https://repositorio.ufmg.br/bitstream/1843/33546/2/license_rdf https://repositorio.ufmg.br/bitstream/1843/33546/3/license.txt |
bitstream.checksum.fl_str_mv |
e768061ee5aa432d1820807855a2fe60 cfd6801dba008cb6adbd9838b81582ab 34badce4be7e31e3adb4575ae96af679 |
bitstream.checksumAlgorithm.fl_str_mv |
MD5 MD5 MD5 |
repository.name.fl_str_mv |
Repositório Institucional da UFMG - Universidade Federal de Minas Gerais (UFMG) |
repository.mail.fl_str_mv |
|
_version_ |
1803589258763567104 |