All Times EDT
Keywords: Bayesian, reference, priors, informative
Much attention has been paid to non-informative Bayesian priors. However, one of the most difficult and important aspects of a Bayesian analysis can be specifying an informative prior. For example in a clinical trial borrowing from historical data, how informative should the prior be? As another example, in an engineering bench test, the prior should reflect the confidence in the engineering design itself. How best to do this? In general, how should we specify mildly supportive, moderately supportive and strongly supportive priors in any given situation? Are there any methods or reference priors that would help us do this?