ABSTRACT
This chapter describes the basic terminology introduced by Rubin and explains in detail in Little and Rubin regarding types of missing data. It illustrates different scenarios of dropout and intermittent missing observations and provides an overview of the potential effects of the different types of missing data. The chapter mentions simple substitution methods and points out their shortcomings in handling missing data. It focuses on the state-of-the-art methods for missing data; namely, multiple imputation (MI), full information maximum likelihood (FIML), and inverse probability weighting methods. The chapter also describes methods for analysis when data are informatively missing and provides references for interested readers. It illustrates the methods on data examples and general guidelines for handling missing data in studies with longitudinal and clustered data. The chapter distinguishes between missing data on predictor and repeatedly measured outcome variables and emphasizes available methods for dealing with both situations.
