ABSTRACT
This chapter discusses prior effective sample size (ESS), which provides a simple, quantitative tool to evaluate how informative a prior actual is before using it. One of the most common criticisms of Bayesian statistics is that inferences depend on the assumed prior, and that this prior is subjective. Criticizing Bayesian statistics on the grounds that priors are subjective is silly, simply because all statistical models and methods rely on subjective assumptions. The pseudo sampling algorithm is an alternative approach to computing the prior model parameters without relying on numerical optimization. It requires that enough prior information be elicited to specify a consistent state of nature, which is called the prior scenario. In addition to the much easier computation, the pseudo sampling approach has other advantages over optimization-based methods for computing priors. Specifying and sampling a prior scenario is a natural and flexible way to incorporate many types of prior information about the state of nature.
