Nonparametric Bayesian robustness

A new, nonparametric, approach to Bayesian robustness is presented. Whereas many
studies in Bayesian robustness have dealt with a parametric sampling distribution, considering
classes of prior distributions on the parameters, here we assume that the sampling
distribution comes from a Dirichlet process with a parameter = , with > 0
and being a probability measure, specied with uncertainty.