ABSTRACT
This chapter serves as a good starting point for newer and more complex modeling approaches. It considers Bayesian inference for model. An important problem in Bayesian analysis is how to define the prior distribution. If prior information about the parameters is available, it should be incorporated in prior distribution. Summaries of these posterior distributions include posterior means and 95% credible intervals, which can be used as Bayesian alternatives to the maximum likelihood estimates and 95% confidence intervals, respectively. The integrated nested Laplace approximations (INLA) library includes a set of functions to operate on marginal distributions. The commonly used functions include inla. dmarginal, inla. pmarginal, inla. qmarginal, inla. mmarginal, and inla. The INLA program allows the user to change the prior for the regression parameters. The model selection in frequentist analysis is commonly based on Akaike information criterion, a maximum likelihood estimate (MLE)-based criterion.
