Applied Multiple Imputation
Advantages, Pitfalls, New Developments and Applications in R
Kristian Kleinke, Jost Reinecke, Daniel Salfrán & Martin Spiess
About the book
Incomplete data are the normal case, not the exception, and multiple imputation is the state-of-the-art answer to them. The standard imputation models, however, assume a world of normally distributed, homoscedastic, comfortably sized data — and a good deal of real research data is none of those things.
This book gives a detailed and readable introduction to multiple imputation, covering both the theory and the practical work of doing it. It discusses the advantages and drawbacks of the various techniques, including imputation quality diagnostics, a topic practitioners need and textbooks often skip. It then turns to current research and to methods designed precisely for the situations in which classical solutions break down: count data, skewed and heteroscedastic variables, clustered and panel structures, and small samples.
Throughout, the presentation stays close to the software. Tutorials cover the widely used R packages norm, pan and mice, and the accompanying code and data sets let readers reproduce every example.
The book addresses social and health scientists and other quantitative researchers working with incomplete data, as well as master’s and doctoral students with a solid basic knowledge of statistics.
Chapters
- Introduction and Basic Concepts
- Missing Data Mechanism and Ignorability
- Missing Data Methods
- Multiple Imputation: Theory
- Multiple Imputation: Application
- Multiple Imputation: New Developments
Chapter 6 describes version 2 of the countimp package in detail. The package has moved on since: version 3 carries its own FCS engine, a formula interface and four further model families, and has a manual of its own.
Citation
Kleinke, K., Reinecke, J., Salfrán, D., & Spiess, M. (2020). Applied multiple imputation: Advantages, pitfalls, new developments and applications in R. Springer. https://doi.org/10.1007/978-3-030-38164-6