Qi Wang, Washington University in Saint Louis

April 25, 2018 - 9:15am to 10:45am

Cupples I, Room 199

*Abstract: Bernstein-von Mises Theorem proves the asymptotic normality of a Bayesian posterior when the models are carefully defined. This presentation mainly focuses on an unspecified version and semiparametric version of the theorem. First, I will briefly introduce the basic Bernstein-von Mises theorem for smooth finite dimensional models. Then, a generalized version which allows misspecification is discussed and compared with the original form. Finally, an infinite-dimensional nuisance parameter will be added to the model, which leads to the discussion of the semiparametric Bernstein-von Mises theorem.*

*Host: Jose Figueroa-Lopez*