Dropouts in longitudinal data analysis occur when some subjects withdraw prematurely from the study. When the processes that govern dropout are related to the measurement process, which is the primary focus of the study, interpretation of the data can be difficult. The Diggle–Kenward model gives a description of the association between the dropout and measurement processes in which completely random, random, or informative dropout models can be obtained by imposing simple and explicit restrictions on the model parameters. Typically, informative dropout parameters are poorly identified, and the associated inferences are therefore sensitive to distributional assumptions.