Bayesian performance evaluation

The distinguishing feature of the Bayesian approach in performance evaluation is that parameter uncertainty of the various models is formally taken into account along with model uncertainty as well. The usual sampling-theory approach proceeds conditionally on the parameter estimates that have been obtained. Of course, the bootstrap can be used but the justification of the bootstrap itself is only asymptotic. We do not intend here to compare and contrast, in detail, the Bayesian versus the sampling-theory approach. We should mention, however, that the Bayesian approach is equipped with numerical techniques that can handle complicated models of performance, where the sampling-theory approach is quite difficult to implement.