chapter  1
Introduction and Examples
Pages 16

This book describes statistical methodology for joint modeling of longitudinal data and time-to-event data. Although our examples focus mostly on biomedical applications, the statistical methods we shall present are applicable to longitudinal follow-up studies in all disciplines. In the area of longitudinal data analysis, joint models were originally developed to address such issues as nonignorable missing data and informative visit times. Missing data are nonignorable when the probability of missingness is related to the missing, unobserved values; otherwise, if the probability of missingness is not related to the missing values, the missing data mechanism is ignorable. Formal definitions of the missing data mechanisms are given in Chapter 2, Section 2.2. Joint models were also studied in the area of time-to-event data analysis for Cox’s (1972) proportional hazards model with time-dependent covariates that are measured intermittently and/or subject to measurement error. In addition, joint models are useful in studies where a repeatedly measured biomarker and a clinical time-to-event outcome are used as co-primary outcome variables to evaluate treatment efficacy.