서브메뉴
검색
Bayesian Modeling and Inference for Complex Dependent Non-Gaussian Data
Bayesian Modeling and Inference for Complex Dependent Non-Gaussian Data
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202105142
- ISBN
- 9798293805679
- DDC
- 574
- 서명/저자
- Bayesian Modeling and Inference for Complex Dependent Non-Gaussian Data
- 발행사항
- [Sl] : University of California, Los Angeles, 2025
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2025
- 형태사항
- 185 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-03, Section: B.
- 주기사항
- Advisor: Banerjee, Sudipto.
- 학위논문주기
- Thesis (Ph.D.)--University of California, Los Angeles, 2025.
- 초록/해제
- 요약This thesis develops novel methodology and software for efficient Bayesian inference in spatial-temporal models, with a focus on non-Gaussian data and misaligned observations, using stacking of predictive densities. Traditional Bayesian approaches for spatial-temporal regression models rely heavily on Markov chain Monte Carlo (MCMC) algorithms, which often face serious convergence issues due to the presence of weakly identified parameters in the spatial-temporal covariance kernel and computational bottlenecks arising from high-dimensional latent processes. To address these challenges, we develop a suite of methods based on predictive stacking, which aggregates analytically tractable posterior distributions conditional on fixed values of certain model parameters. By extending the Diaconis-Ylvisaker conjugate prior framework and leveraging generalized conjugate multivariate distribution theory, we enable exact conditional inference and circumvent the need for iterative algorithms in the non-Gaussian setup. These methodologies are implemented in the R package spStack, which offers a fast, parallelizable, and user-friendly tool for Bayesian geostatistical modeling. We apply our methods to various real-world problems, including the analysis of bird count data and spatially-temporally misaligned air pollution and asthma-related hospital visits data. The latter setting, commonly referred to as the change of support problem, arises frequently in environmental health studies where pollutant exposures are recorded at fine spatial-temporal resolutions, while health outcomes are aggregated over coarser units due to survey or privacy constraints. We pursue predictive stacking in this context to synthesize inference across disparate spatial-temporal supports, demonstrating competitive predictive accuracy and computational efficiency over traditional MCMC-based approaches. In addition to the main line of research on predictive stacking, this thesis also includes brief work on Bayesian modeling of ordinary differential equations, broadening the scope of its methodological contributions. Collectively, the work presented here offers scalable Bayesian alternatives for spatial-temporal data analysis, expanding the practical utility of hierarchical models in the environmental and health sciences.
- 일반주제명
- Biostatistics
- 일반주제명
- Statistics
- 일반주제명
- Information technology
- 키워드
- Conjugate prior
- 키워드
- Geostatistics
- 기타저자
- University of California, Los Angeles Biostatistics 0132
- 기본자료저록
- Dissertations Abstracts International. 87-03B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2025 us c eng d■001000017359586
■00520260202105142
■006m o d
■007cr#unu||||||||
■020 ▼a9798293805679
■035 ▼a(MiAaPQ)AAI32240768
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a574
■1001 ▼aPan, Soumyakanti.
■24510▼aBayesian Modeling and Inference for Complex Dependent Non-Gaussian Data
■260 ▼a[Sl]▼bUniversity of California, Los Angeles▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a185 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-03, Section: B.
■500 ▼aAdvisor: Banerjee, Sudipto.
■5021 ▼aThesis (Ph.D.)--University of California, Los Angeles, 2025.
■520 ▼aThis thesis develops novel methodology and software for efficient Bayesian inference in spatial-temporal models, with a focus on non-Gaussian data and misaligned observations, using stacking of predictive densities. Traditional Bayesian approaches for spatial-temporal regression models rely heavily on Markov chain Monte Carlo (MCMC) algorithms, which often face serious convergence issues due to the presence of weakly identified parameters in the spatial-temporal covariance kernel and computational bottlenecks arising from high-dimensional latent processes. To address these challenges, we develop a suite of methods based on predictive stacking, which aggregates analytically tractable posterior distributions conditional on fixed values of certain model parameters. By extending the Diaconis-Ylvisaker conjugate prior framework and leveraging generalized conjugate multivariate distribution theory, we enable exact conditional inference and circumvent the need for iterative algorithms in the non-Gaussian setup. These methodologies are implemented in the R package spStack, which offers a fast, parallelizable, and user-friendly tool for Bayesian geostatistical modeling. We apply our methods to various real-world problems, including the analysis of bird count data and spatially-temporally misaligned air pollution and asthma-related hospital visits data. The latter setting, commonly referred to as the change of support problem, arises frequently in environmental health studies where pollutant exposures are recorded at fine spatial-temporal resolutions, while health outcomes are aggregated over coarser units due to survey or privacy constraints. We pursue predictive stacking in this context to synthesize inference across disparate spatial-temporal supports, demonstrating competitive predictive accuracy and computational efficiency over traditional MCMC-based approaches. In addition to the main line of research on predictive stacking, this thesis also includes brief work on Bayesian modeling of ordinary differential equations, broadening the scope of its methodological contributions. Collectively, the work presented here offers scalable Bayesian alternatives for spatial-temporal data analysis, expanding the practical utility of hierarchical models in the environmental and health sciences.
■590 ▼aSchool code: 0031.
■650 4▼aBiostatistics
■650 4▼aStatistics
■650 4▼aInformation technology
■653 ▼aConjugate prior
■653 ▼aExponential family
■653 ▼aGeostatistics
■653 ▼aMechanistic systems
■653 ▼aModel combination
■653 ▼aPredictive modeling
■690 ▼a0308
■690 ▼a0489
■690 ▼a0463
■71020▼aUniversity of California, Los Angeles▼bBiostatistics 0132.
■7730 ▼tDissertations Abstracts International▼g87-03B.
■790 ▼a0031
■791 ▼aPh.D.
■792 ▼a2025
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17359586▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
