Please join us for this CCASA Tuesday Luncheon program.
Extended Abstract
This talk will provide an overview of Bayesian nonparametric modeling, along with an illustrative application. This approach to statistical modeling is conditioned on a “nonparametric” prior distribution that gives positive support to the set of all sampling distributions {F}. Furthermore, under mild conditions involving the specification of the nonparametric prior distribution, posterior consistency is guaranteed, in the sense that the posterior distributions accumulate around the true distribution F_0 with probability one. In this sense, a Bayesian nonparametric model represents a true model for a given set of data (generated by a true distribution F_0).
It will also be shown that a nonparametric prior (that leads to posterior consistency) guarantees coherent statistical modeling, i.e., modeling that is consistent with the axioms of quantitative coherence. Furthermore, in passing, it will be demonstrated that under the Bayesian decision-theoretic approach to the point-estimation problem, the maximum likelihood estimate of a parameter arises as the optimal choice of point-estimate under a “noninformative” nonparametric prior.
To make a reservation, Contact
Lou Fogg, VP for Luncheons
Phone 312-942-6239
E-mail: louis_fogg@rush.edu
Cost will be $30 for members and $35 for non- members. Nonmembers, join the chapter for a year for only $15 and get the discount plus all the benefits of membership! Visa and Mastercard accepted.