ABSTRACT
As shown in Chapter 7, survival data often arise in longitudinal studies. In practice, we may need to model a time-to-event process and a longitudinal process jointly. A common situation is in survival models with time-dependent covariates, in which covariate data may be missing at failure times, so a longitudinal model for the covariates is required to address missing covariates or covariate measurement errors. As another example, in modeling longitudinal data with dropouts, we may also need to model the times to dropout or times to death, with the objectives of avoiding possible biases in the estimation of the longitudinal model, as well as studying the association between the time to event and characteristics of the longitudinal trajectories such as initial slopes or intercepts. In both cases, joint modeling of the longitudinal data and the survival data is required.
