Joint Modelling of Survival and Longitudinal Data

描述(申请人提供)：在实验性衰老研究和其他纵向研究中，观察一个感兴趣的事件时间，称为生存时间，以及在几个时间点测量的纵向协变量，已经变得越来越常见。科学界越来越感兴趣的是同时对这两个过程进行建模，以探索它们之间的关系，并从模型建立过程中的每个组成部分借用力量。 由于计算环境的迅速改善，这种联合建模方法已经变得可行。在过去的十年里，已经提出了相当多的创新方法，但它们通常涉及到严格的模型限制。这项提议旨在开发限制较少的替代模型，并针对相关的计算和理论挑战。我们将发展： 1.提供模型灵活性的纵向协变量的非参数混合效应方法； 2.更通用的生存模型和方法，以及生存部分的模型检测工具； 3.减少计算负担和提供计算稳定性的筛分方法； 4.用户友好的软件，包括AIMS 1-3在老龄化研究中的应用； 5.结合不可忽略的审查和各种测量误差结构的附加工具。 这项研究是由几项老龄化研究推动的，这些研究解决了生殖模式和寿命之间的关系。它从称为功能数据分析的相关领域提取工具，将观察到的纵向数据视为平滑的底层过程的分散实现，可能观察到有测量误差。新方法不仅将阐明寿命与生育史等纵向指标之间的关系，而且还将广泛适用于临床和流行病学研究。它们涉及新出现的统计工具，这些工具将提供先进的方法和灵活的方法来模拟复杂的生物系统，并将促进模型检查。

DESCRIPTION (provided by applicant): In experimental aging studies and other longitudinal studies, it has become increasingly common to observe an event time of interest, called survival time, along with longitudinal covariates measured at several time points. A growing interest in the scientific community is to model both processes simultaneously to explore their relationship and to borrow strength from each component in the model building process. Such joint modeling approaches have become feasible due to the rapidly improving computing environment. Quite a few innovative approaches have been proposed in the last decade, but they typically involve strong model restrictions. This proposal aims at developing less restrictive alternative models and targets the associated computational and theoretical challenges. We will develop: 1. Nonparametric mixed-effects approach for longitudinal covariates that provide model flexibility; 2. More general survival models and approaches, and model checking tools for the survival component; 3. The method of sieves to alleviate the computational burden and to provide computational stability; 4. User-friendly software including applications of aims 1-3 to aging studies; 5. Additional tools to incorporate non-ignorable censoring and various measurement error structures. This research is motivated by several aging studies that address the relationship between patterns of reproduction and longevity. It draws tools from a related area called Functional Data Analysis, viewing the observed longitudinal data as scattered realizations of a smooth underlying process, possibly observed with measurement errors. The new approaches will not only shed light on the relationship between life-span and longitudinal markers such as reproductive histories, but will also be broadly applicable to clinical and epidemiological studies. They involve emerging statistical tools that will provide advanced methodology and flexible approaches to model complex biological systems, and will also facilitate model checking.