Comments, with reply, on "Reliability of modified designs-a Bayes analysis of an accelerated test of electronic assemblies" by L. Hart

Abstract

In examining the above-named work (See ibid., vol.39, June 1990, p.140-4) of L. Hart, the commenter (Ganter) observes that Hart's choice of a beta prior is reasonable and convenient only because it is well known that a beta prior yields a beta posterior in the Bayes method. It is argued that, because the binomial is exact, Hart can have no real justification for his beta prior from his A test results. Hart presents a reply. In addition, H.F. Martz and R. Fisher comment on the above exchange of views. They conclude that Ganter's claim that Hart achieved no inference gain in using Bayes methods is unfounded. They argue that Ganter's conclusion depends on the particular data set chosen and that his choice of ratio is inappropriate. >