ABSTRACT
Missing data are a common problem in longitudinal data sets, as the overview in Chapter 17 discussed. This chapter considers likelihood-based methods for handling this problem, based on parametric models for the data and missing-data mechanism. These models can also form the basis for multiple imputation approaches discussed in Chapter 21. Approaches based on estimating equations other than the likelihood, including inverse probability weighting methods, are discussed in Chapter 20. A useful tutorial that discusses both likelihood-based and estimating equations approaches is Hogan, Roy, and Korkontzelou (2004).
