Composite likelihoods in the Bayesian inference: an application

While the likelihood function plays a central role in the theory of statistical inference for parametric models, there are situations where modifications of the likelihood are needed, for example for robustness or for the complexity of the full likelihood. This contribution is concerned with a special form of pseudo-likelihood useful when complex interdependencies are involved in the full likelihood. In these situations, the idea is to use approximate likelihoods based, for example, on bivariate marginal distributions.