Statistical Methods for Analyzing Interval-Censored Multi-State Data- [electronic resource]
Statistical Methods for Analyzing Interval-Censored Multi-State Data- [electronic resource]
- 자료유형
- 학위논문파일 국외
- 최종처리일시
- 20240214101233
- ISBN
- 9798380133081
- DDC
- 574
- 저자명
- Gu, Yu.
- 서명/저자
- Statistical Methods for Analyzing Interval-Censored Multi-State Data - [electronic resource]
- 발행사항
- [S.l.]: : The University of North Carolina at Chapel Hill., 2023
- 발행사항
- Ann Arbor : : ProQuest Dissertations & Theses,, 2023
- 형태사항
- 1 online resource(174 p.)
- 주기사항
- Source: Dissertations Abstracts International, Volume: 85-02, Section: B.
- 주기사항
- Advisor: Lin, Danyu;Zeng, Donglin.
- 학위논문주기
- Thesis (Ph.D.)--The University of North Carolina at Chapel Hill, 2023.
- 사용제한주기
- This item must not be sold to any third party vendors.
- 초록/해제
- 요약Interval-censored multi-state data are commonly encountered in studies of chronic diseases, where a subject's health status is characterized by a finite number of disease states and transitions between states are only known to occur within broad time intervals. The fact that none of the transition times are directly observed makes regression analysis extremely challenging, both theoretically and computationally. This dissertation develops novel statistical methods for interval-censored multi-state data analysis in three projects.In the first project, we provide a new framework based on semiparametric proportional intensity models to analyze general interval-censored multi-state data. We adopt nonparametric maximum likelihood estimation (NPMLE) and develop a stable EM algorithm based on latent Poisson random variables. We establish a rigorous asymptotic theory for the resulting estimators and demonstrate the satisfactory performance of the proposed inference procedures through extensive simulation studies. Finally, we apply these methods to the Atherosclerosis Risk in Communities study.The second project extends the first project by incorporating an absorbing state, whose entry time is exactly known or right-censored, while the transient state at the previous instant remains unknown. We combine NPMLE with sieve estimation for inference and devise a stable EM algorithm for computation. We establish the asymptotic properties of the proposed estimators using novel theoretical arguments. Furthermore, we perform dynamic prediction of future disease process based on the evolving process history. Finally, we assess the performance of the proposed methods through simulation studies and illustrate their application using the cardiac allograft vasculopathy data.In the third project, we introduce a novel framework based on transition models with random effects for assessing treatment effects in COVID-19 phase-3 clinical trials. Our models estimate treatment effects on the odds of transitions between severity categories, using the entire clinical course without the need for missing data imputation. We discuss flexible model parameterizations for various applications. We also demonstrate how fitted transition models can be integrated with G-computation to examine complex endpoints and increase the power for detecting treatment effects. Finally, we evaluate the proposed methods through simulation studies and apply them to data from the Adaptive COVID-19 Treatment Trial.
- 일반주제명
- Biostatistics.
- 일반주제명
- Statistics.
- 일반주제명
- Public health.
- 키워드
- EM algorithm
- 키워드
- Random effects
- 기타저자
- The University of North Carolina at Chapel Hill Biostatistics
- 기본자료저록
- Dissertations Abstracts International. 85-02B.
- 기본자료저록
- Dissertation Abstract International
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
detalle info
- Reserva
- No existe
- Mi carpeta
- Primera solicitud
- 비도서대출신청
- 야간 도서대출신청
로그인 후 이용 가능합니다.