Discussion of "Specifying Prior Distributions in Reliability Applications"

This paper provides a contribution to the discussion on the paper of Tian et al. (2023) entitled "Specifying Prior Distributions in Reliability Applications" regarding the specification of prior distributions on the parameters of the log-location-scale distributions, such as the Weibull and the lognormal distribution. The use of several combinations of the suggested prior distributions is here suggested to correctly perform a model selection (among the log-location-scale family of distributions) based on the Bayes factor, that is in compliance with the requirements for prior distributions: a) all the prior distributions must be proper, b) the prior distribution on a parameters indexing only one of the competing models must be noninformative or weakly informative, and c) a strong prior distribution can be used only on a parameter indexing both the competing models, in their original or new parameterization.