Discussion of: A Bayesian information criterion forsingular models

The BIC can be viewed as an easily computable proxy to fully Bayesian model choice, which is conducted by comparing the marginal likelihood (or evidence) for each of the models. However, the derivation for BIC relies on informative data and a noninformative prior and that the models under consideration are non-singular. Thus the development of associated information criteria that are suitable when the models are singular is an important research problem. Hence, the authors should be congratulated for their contribution...