
4
Bayesian analysis in practice
4.1 Introduction
As discussed in Chapter 2, if it is possible to sample from the full conditional
distributions, Gibbs sampling algorithms can be programmed in a relatively
easy way. However, in most practical situations, the required conditional dis-
tributions are not analytically tractable and therefore it is necessary to ap-
proximate them (e.g. by means of algorithms such as Metropolis-Hastings or
slice sampling) before Gibbs sampling can be performed.
The most popular software that allows the semi-automatisation of the
MCMC procedures is BUGS, and particularly its MS Windows incarnations
WinBUGS (Spiegelhalter et al., ...