New Tools for Consistency in Bayesian Nonparametrics
Gabriella Salinetti · 2003
Abstract Consistency in Bayesian nonparametric statistical problems continues to register increasing attention, motivated by various and easy to share reasons, widely illustrated in the rich literature on the subject starting from Diaconis and Freedman (1986) to the recent Ghosh and Ramamoorthi (2002). The difficulty in nonparametrics is in the dimension of the problem: the “parameter” itself is typically a probability measure, the prior is a probability measure on a space of probability measures, and the resulting posterior, after seeing the data, is a random probability measure.