서브메뉴
검색
Targeted Learning in Heterogeneous Effect Estimation and Adaptive Group Sequential Design
Targeted Learning in Heterogeneous Effect Estimation and Adaptive Group Sequential Design
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20250211151455
- ISBN
- 9798384449393
- DDC
- 574
- 저자명
- Li, Haodong.
- 서명/저자
- Targeted Learning in Heterogeneous Effect Estimation and Adaptive Group Sequential Design
- 발행사항
- [Sl] : University of California, Berkeley, 2024
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2024
- 형태사항
- 89 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 86-04, Section: B.
- 주기사항
- Advisor: Hubbard, Alan E.;van der Laan, Mark J.
- 학위논문주기
- Thesis (Ph.D.)--University of California, Berkeley, 2024.
- 초록/해제
- 요약Targeted learning represents a general statistical framework that connects different fields such as machine learning and causal inference. It is often used to estimate causal effects in both experimental and observational studies. There are three important tools in this framework: 1) targeted maximum likelihood estimation (TMLE), 2) super learning (SL) and 3) highly adaptive lasso (HAL). Among them, TMLE is the core estimation methodology; SL is an ensemble learning algorithm which can serve for the initial estimation step of TMLE; HAL is a powerful nonparametric loss-based estimator which can be included into a SL library as a candidate learner.This thesis comprises three cases studies demonstrating the method advancement and applications of these targeted learning tools in treatment heterogeneity, adaptive design and realistic simulations. Chapter 1 describes a new TMLE for a treatment effect variable importance parameter. This target parameter can be used to identify potential driven factors for treatment heterogeneity. A robust estimate reveals details on the importance of different covariates and how they lead to different treatment impacts. Chapter 2 introduces a SL algorithm that aims to select the best treatment policy at each time point in an adaptive group sequential design by maximizing the average cumulative reward. In chapter 3, we benchmark multiple widely used methods for estimation of the average treatment effect using ten different nutrition intervention studies data. A nonparametric regression method, undersmoothed HAL, is used to generate the simulated distribution which preserves important features from the observed data and reproduces a set of true target parameters.
- 일반주제명
- Biostatistics
- 일반주제명
- Medicine
- 일반주제명
- Computer science
- 키워드
- Super learning
- 키워드
- Machine learning
- 기타저자
- University of California, Berkeley Biostatistics
- 기본자료저록
- Dissertations Abstracts International. 86-04B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008250123s2024 us c eng d■001000017161864
■00520250211151455
■006m o d
■007cr#unu||||||||
■020 ▼a9798384449393
■035 ▼a(MiAaPQ)AAI31297035
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a574
■1001 ▼aLi, Haodong.
■24510▼aTargeted Learning in Heterogeneous Effect Estimation and Adaptive Group Sequential Design
■260 ▼a[Sl]▼bUniversity of California, Berkeley▼c2024
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2024
■300 ▼a89 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 86-04, Section: B.
■500 ▼aAdvisor: Hubbard, Alan E.;van der Laan, Mark J.
■5021 ▼aThesis (Ph.D.)--University of California, Berkeley, 2024.
■520 ▼aTargeted learning represents a general statistical framework that connects different fields such as machine learning and causal inference. It is often used to estimate causal effects in both experimental and observational studies. There are three important tools in this framework: 1) targeted maximum likelihood estimation (TMLE), 2) super learning (SL) and 3) highly adaptive lasso (HAL). Among them, TMLE is the core estimation methodology; SL is an ensemble learning algorithm which can serve for the initial estimation step of TMLE; HAL is a powerful nonparametric loss-based estimator which can be included into a SL library as a candidate learner.This thesis comprises three cases studies demonstrating the method advancement and applications of these targeted learning tools in treatment heterogeneity, adaptive design and realistic simulations. Chapter 1 describes a new TMLE for a treatment effect variable importance parameter. This target parameter can be used to identify potential driven factors for treatment heterogeneity. A robust estimate reveals details on the importance of different covariates and how they lead to different treatment impacts. Chapter 2 introduces a SL algorithm that aims to select the best treatment policy at each time point in an adaptive group sequential design by maximizing the average cumulative reward. In chapter 3, we benchmark multiple widely used methods for estimation of the average treatment effect using ten different nutrition intervention studies data. A nonparametric regression method, undersmoothed HAL, is used to generate the simulated distribution which preserves important features from the observed data and reproduces a set of true target parameters.
■590 ▼aSchool code: 0028.
■650 4▼aBiostatistics
■650 4▼aMedicine
■650 4▼aComputer science
■653 ▼aHighly adaptive lasso
■653 ▼aSuper learning
■653 ▼aTargeted maximum likelihood estimation
■653 ▼aMachine learning
■690 ▼a0308
■690 ▼a0984
■690 ▼a0564
■71020▼aUniversity of California, Berkeley▼bBiostatistics.
■7730 ▼tDissertations Abstracts International▼g86-04B.
■790 ▼a0028
■791 ▼aPh.D.
■792 ▼a2024
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17161864▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
