Approaches for the joint evaluation of hypothesis tests: Classical testing, Bayes testing, and joint confirmation

摘要

The occurrence of decision problems with changing roles of null and alternative hypotheses has increased interest in extending the classical hypothesis testing setup. Particularly, confirmation analysis has been in the focus of some recent contributions in econometrics. We emphasize that confirmation analysis is grounded in classical testing and should be contrasted with the Bayesian approach. Differences across the three approaches - traditional classical testing, Bayes testing, joint confirmation - are highlighted for a popular testing problem. A decision is searched for the existence of a unit root in a time-series process on the basis of two tests. One of them has the existence of a unit root as its null hypothesis and its non-existence as its alternative, while the roles of null and alternative are reversed for the other hypothesis test.