ABSTRACT
For the mixed effects models considered in previous chapters, parametric models are often assumed for the response, for the random effects, for the incompletely observed covariates, and for the missing data or measurement error models. So the joint models often contain too many parameters, and this may lead to poor estimation of the main parameters or may lead to identifiability problems, especially if the observed data are not rich. Bayesian methods offer the advantage of borrowing information from similar studies or from experts, which are then incorporated in the current analysis in the forms of prior distributions for the parameters. Such prior information helps estimating parameters that may be poorly identified by the current observed data alone.
