ABSTRACT
This chapter considers the rgeneric latent effect that allows new latent effects implemented in R code. It discusses how to combine integrated nested Laplace approximation (INLA) and Markov chain Monte Carlo to increase the number of models that can be fitted using INLA. Specifying the latent effect as a Gaussian Markov random fields also requires a binary representation of the precision matrix to exploit conditional independence properties. i.e., a ‘graph’. R. Bivand et al. describe a novel approach to fit new models with INLA. Their approach is based on conditioning on one of the hyperparameters in the model and fit conditional models on several values of this parameter. When INLA is used to fit several models in parallel on the same machine it is convenient to set the number of threads used by each INLA process to a value that ensures that no more cores than available are used.
