ABSTRACT
The integrated nested Laplace approximation (INLA) approach is implemented in the R package R-INLA. The package also provides estimates of different criteria to assess and compare Bayesian models. These include the model deviance information criterion, the Watanabe-Akaike information criterion, the marginal likelihood, and the conditional predictive ordinates. The syntax of the linear predictor in R-INLA is similar to the syntax used to fit linear models with the function. A control predictor list with the specification of several predictor variables such as link which is the link function of the model, and compute which is a Boolean variable that indicates whether the marginal densities for the linear predictor should be computed. The values of the priors can be changed in the control. A summary linear predictor data frame with the mean, standard deviation, and quantiles of the linear predictors.
