Flexible Imputation of Missing Data (2nd Edition)
Abdolvahab Khademi
Abstract
Occurrence of missing data can cause serious issues, including decreased sample size, biased estimates, and algorithmic problems.Therefore, proper treatment of missing data is a significant part of data analysis in statistics, especially in clinical and experimental studies.Treatment of missing data is usually included in a section of its own in most textbooks, presenting best or most convenient practices, based on the methods presented and the statistical sophistication of the audience.However, the contexts, complexity, and severity of missing data are so diverse and complicated that its treatment warrants a volume of its own.There are currently several books on missing data, ranging from practical to more theoretical.Flexible Imputation of Missing Data (2nd Edition) is an updated addition to the literature on missing data, which combines practice, theory, and applications using the R programming language.The twelve chapters of the book are grouped in four sections: the basics, advanced techniques, case studies, and extensions.Exposition in each chapter is accompanied by plenty of graphs, code, examples, and exercises.The code and the entire book are available online at the author's personal website.The programming language of the book is R and the treatment of missing data is performed by the package mice, which was created by the author.Chapter 1, introduction, provides motivation for dealing with missing data, how missing data occur (e.g. by nonresponse or attrition), current practice in the literature, categories of missing data (MCAR, MAR, MNAR), and common fixes and their advantages and drawbacks.Common practices, such as listwise and pairwise deletion methods, mean imputation, (stochastic) regression regression, last/baseline observation carried forward (LOCF and BOCF), and indicator method (popular in public health) are discussed.The largest portion of the chapter is dedicated to multiple imputation.According to the author, the emphasis of the entire book is on multiple imputation because of its efficiency and statistical properties compared to other methods.The entire Chapter 2, multiple imputation, is devoted to the method of multiple imputation (MI), starting with a historical sketch of the concept and practice of MI.