Year of Publication
Thompson, DJS; El-Shaarawi, A.H.; Piegorsch, W.
Encyclopedia of Environmetrics
John Wiley & Sons, Ltd.
conditional independence model; conditional model; hierarchical model; joint model; mixed outcomes
Joint modeling encompasses strategies to simultaneously model several outcomes of interest. There are three principal strategies; classical joint modeling, conditional models, and conditional independence models. Likely the most pervasive area of joint modeling is in the modeling of longitudinal and time-to-event data; in particular, accounting for drop-out in longitudinal data or incorporating error-prone, sporadically measured, longitudinal outcomes in models for event times. Conditional independence is a popular strategy, which assumes the outcomes of interest are noisy, independent measures of some underlying latent process; it is this process that induces their correlation providing a tractable assumption in many practical settings.