Posterior predictive checks: principles and discussion procedures (Q1424595)

This paper presents a brief overview of the recent development and ways of application of the posterior predictive check (PPC) approach for detecting departures between the data and the posited model. It covers the following topics: basic principles of the PPC approach, PPC using tests statistics and discrepancy measures, practical estimations of posterior predictive \(p\)-values, interpolation and frequency properties of p-values, and power of posterior predictive tests. The authors compare PPC-procedures with the classical approach to goodness-of-fit testing and outline the following advantages of the PPC-procedures: they do not rely on asymptotic properties, are easy to perform and useful for model monitoring.