Computation of reference Bayesian inference for variance components in longitudinal studies (Q626239)

A generalized linear model is considered with Gaussian random effects. Baysian estimation is discussed of the effects covariance matrix \(D\). The authors describe approximate Jeffreys' priors for \(D\) and compare Bayesian, restricted maximum likelihood and penalized quasi likelihood estimates from the frequentist point of view. Results of simulations and applications to biomedical real life data are presented. Some computational issues of Markov chain Monte Carlo (MCMC) procedures for sampling from the posterior are discussed.