There are times when, even as a Bayesian, one is interested in calibration. Model checking without a specified alternative is an example. Frequentist ideas --

There are times when, even as a Bayesian, one is interested in calibration. Model checking without a specified alternative is an example. Frequentist ideas -- sampling from the model and comparing it to the observed distribution -- can be helpful here. I'm thinking of Rubin (1984): http://www.cs.princeton.edu/courses/archive/fall11/cos597C/reading/Rubin1984.pdf http://www.cs.princeton.edu/courses/archive/fall11/cos597C/r...