An evaluation of alternative methods for testing hypotheses, from the perspective of Harold Jeffreys

Our original article provided a relatively detailed summary of Harold Jeffreys's philosophy on statistical hypothesis testing. In response, Robert (2016) maintains that Bayes factors have a number of serious shortcomings. These shortcomings, Robert argues, may be addressed by an alternative approach that conceptualizes model selection as parameter estimation in a mixture model. In a second comment, Chandramouli and Shiffrin (2016) seek to extend Jeffreys's framework by also taking into consideration data distributions that do not originate from either of the models under test. In this rejoinder we argue that Robert's (2016) alternative view on testing has more in common with Jeffreys's Bayes factor than he suggests, as they share the same “shortcomings”. On the other hand, we show that the proposition of Chandramouli and Shiffrin (2016) to extend the Bayes factor is in fact further removed from Jeffreys's view on testing than the authors suggest. By elaborating on these points, we hope to clarify our case for Jeffreys's Bayes factors.