Applied Multiple Imputation
Applied Multiple Imputation
复制标题
应用多重插补
DOI：
10.1007/978-3-030-38164-6
复制
发表时间：
2020
期刊：
影响因子：
--
通讯作者：
Spiess
中科院分区：
文献类型：
--
作者：
Kleinke;Reinecke;Salfrán;Spiess
经验数据很少被完全观察到。如何充分分析受缺失值影响的数据集通常不是学士或硕士课程的重点，因此应用研究人员经常采用简单的临时解决方案，如完整的案例分析或（无条件）平均插补，在前一种情况下通常是不合适的，在后一种情况下几乎总是不合适的。博士生和应用研究人员经常告诉我们，他们读过关于多重插补（MI）的文章，但他们发现以自学的方式学习MI的应用很麻烦。本书将在这方面提供有用的教程。但这不仅仅是如何应用多元智能技术的问题。了解何时应用MI也很重要--在哪些场景中适合使用该方法，以及何时不适合。为此，应用研究人员至少应该意识到缺失值可能导致的问题，并对多重插补理论有基本的了解，即MI试图纠正的内容和潜在的假设是什么，写这本书背后的意图是满足两个目的，即对基本概念和统计基础提供详细的介绍，并在理论基础上论证了多元智能的实际应用。多重插补方法背后的统计概念将在第三章中解释。1-4.每一章都包括对理论和各种概念的深入描述和讨论，这些概念是理解MI为什么和何时起作用（或不起作用）所必需的，并使用户能够决定某种方法或软件是否适用于他或她的特定应用。每一个概念都用例子来说明。读者如果只想对这个理论有一个简要的概述，可以在每一章的末尾找到一个总结部分。第5章和第6章描述并解释了如何做到这一点。如果读者一开始对统计背景不感兴趣，他们可以从应用章节开始。5和6，学习如何将这些技术应用于经验数据集，并在需要时回到理论部分（各种交叉引用将引导读者到相应的部分，在那里他们可以找到更多的细节和解释）。在我们的教程中，我们明确地不想使用那些运行良好的好例子。相反，我们的例子
Empirical data are seldom completely observed. How to adequately analyse data sets affected by missing values is usually not the focus of courses at bachelor or master level, and thus applied researchers often resort to simple ad hoc solutions like complete case analysis or (unconditional) mean imputation, which in the former case is often inappropriate and is virtually always inappropriate in the latter case. PhD students and applied researchers often tell us that they have read about multiple imputation (MI) but that they have found it cumbersome to learn the application of MI in an autodidactic way. This book will provide helpful tutorials in this regard. But it is not only the question about how to apply MI techniques. It is also important to know when to apply MI—in which scenarios it is appropriate to use the method, and when it is not. To this end, applied researchers should be at least aware of the possible problems caused by missing values and have a basic understanding of the theory of multiple imputation, that is, what MI tries to correct and what the underlying assumptions are.The intention behind writing this book was to meet both ends, ie to provide a detailed introduction to the basic concepts and statistical underpinnings, and to demonstrate the practical application of MI based on the underlying theory. The statistical concepts behind the method of multiple imputation are explained in Chaps. 1–4. Each of the chapters consists of an in-depth description and discussion of the theory and the various concepts needed to understand why and when MI works (or not) and to enable users to decide whether a certain method or software may work for his or her particular application. Each concept is illustrated with examples. Readers who only want to get a brief overview over the theory may find a summary section at the end of each chapter. Chapters 5 and 6 describe and explain the how. Should readers—at first—not be interested in the statistical background, they can start with the applied Chaps. 5 and 6 to learn how to apply the techniques to empirical data sets and come back to the theory part when needed (various cross-references will direct readers to the respective sections, where they can find further details and explanations). In our tutorials, we explicitly did not want to use nice and well-behaved examples that work and run through nicely without further ado. Instead we base our examples