This paper considers a Bayesian approach to pint null hypothesis testing when the sampling distribution nelongs to a particular class, defined in Gleser and Hwang (1987). Different approaches to the class have been used in the literature; however, none of them is fully satisfactory. A Bayesian viewpoint is adopted to analyze the behaviour of several version of the Bayes factor in this context. We first discuss the use of a proper Bayes factor and we show the high sensitivity of conclusions with respect to prior inputs. Also, alternative approaches based on default Bayes factors are considered in the particular context of linear calibration.
Barbieri, M.M., Liseo, B., Petrella, L. (1998). Bayes factors at work in a challenging class of problems. In Proceedings of the Workshop on Model Selection (pp.109-132). BOLOGNA : Pitagora.