Statistical Methods on Challenging Issues of Longitudinal and Lifetime Data

随着统计研究的进步，应用中出现了更多的挑战，带来了新的问题，使现有的方法失效。标准的统计方法是在收集的数据是“完美的”的假设下发展起来的。然而，这个假设在实践中经常被违背。特别是，测量误差和观测缺失是经常发生的。在数据分析中，测量误差一直是一个长期存在的问题，应该尽可能避免。缺失在应用程序中无处不在，它给标准推理方法带来了巨大的挑战。过去几十年来，在处理测量误差和缺失数据的方法方面取得了相当大的进展，其中大部分工作集中在纵向和终身研究中产生的问题上。越来越多的文献和新出现的词汇有助于提高人们对测量误差或缺失数据可能对治疗和其他协变量效应的推断产生影响的认识。

As statistical research advances, additional challenges are emerging from applications, bringing new issues that make existing methods break down. Standard statistical methods have been developed under the assumption that the collected data are "perfect". This assumption is, however, often violated in practice. In particular, measurement error and missing observations occur commonly. Measurement error has been a long standing concern in data analysis, and it should be avoided whenever possible. Missingness is ubiquitously present in applications, and it creates tremendous challenges to standard inference methods. There have been considerable advances in the past few decades on methodology for dealing with measurement error and missing data, where much of the work has focused on issues arising from longitudinal and lifetime studies. The growing literature and emerging lexicon have helped raise awareness of the impact that measurement error or missing data can have on inferences regarding treatment and other covariate effects.