ABSTRACT
As already discussed in Chapter 2, in fitting stochastic models to data using classical inference we think in general in terms of selecting between models, using tests and information criteria, and then checking how well a selected model describes data. There are also possibilities of model averaging. From a Bayesian perspective there is a natural way to select posterior model probabilities to produce model averaging, though in practice there may be difficulties in implementation, and one can check the fit of particular models using Bayesian p-values.
