The analysis of survival data

流行病学、统计学和预防研究司的研究以及一般生物医学研究对生存数据的分析非常感兴趣，因为遇到的数据通常都是这种性质的。我们在这个项目中的目标是多方面的： (1)开发分析具有非标准类型不完全性的事件的时间的方法，这些方法超出了右删失，如随机截断，当前状态和区间删失。 大量的文献存在处理随机右删失的数据，但是，需要更多的努力来开发方法来处理其他类型的不完全性，经常发生。例如，在许多回顾性妊娠研究中经常观察到随机截断，其中女性仅在研究开始时怀孕才入组研究。 (2)开发处理多变量生存数据的方法。该项目的这一部分是必要的，许多流行病学研究的需要，其中的兴趣不仅仅是在建模一个简单的感兴趣的事件，而是一系列复杂的事件。这一部分的重点是开发方法来分析多变量生存数据，如多阶段数据，复发事件和竞争风险数据。这种多变量数据的例子可以在DESPR进行的各种研究中找到，NICHD如LIFE研究和其他前瞻性妊娠研究，安全劳动仅举几例。在这里开发的方法将解决的问题，如通过不同阶段的劳动进展建模，建模时间怀孕分娩或损失。此外，出于需要更好地了解与反复发生与妊娠丢失相关的不良结局的妇女相关的风险因素，我们打算使用竞争风险数据和复发事件数据来研究此类事件的建模

The analysis of survival data is of considerable interest to studies in Division of Epidemiology, Statistics and Prevention Research and also to biomedical research in general as the data encountered is typically of that nature. Our objective in this project is multifold: (1) Develop methods to analyze time to an event which have non-standard type of incompleteness, that are beyond right censoring, like random truncation, current status and interval censoring. Extensive literature exists for dealing with data which are randomly right censored However, more effort is needed to develop methods to deal with other type of incompleteness which occur frequently. For example, random truncation is frequently observed in many retrospective pregnancy studies where women are enrolled in the study only if they have gotten pregnant by start of study. (2) Develop methods to deal with multivariate survival data. This part of the project is necessitated by needs of many epidemiological studies, where the interest is not just in modeling one simple event of interest, but rather a host of complex events. The focus of this part is to develop methods to analyze multivariate survival data like the multistage data, recurrent events and competing risks data. Examples of such multivariate data can be found in various studies being conducted in DESPR, NICHD like LIFE study and other prospective pregnancy studies, Safe Labor to name few. The methods developed here will address questions like modeling of progression of labot through various stages, modeling time to pregnancy to delivery or loss. Also, motivated by need to better understand the risk factors associated with women who suffer repeated adverse outcomes associated with pregnancy loss, we intend to study modeling such events using both the competing risks data as well as recurrent events data