PLACE: 319 Snedecor
SPEAKER: Dongchu Sun, University of Missouri - Columbia
TITLE: Objective Bayesian Semiparametric Models and Space-Time Interaction
ABSTRACT
Bayesian analysis, especially of semiparametric and hierarchical linear
mixed models, has received much attention recently. There is often not
enough information on hyperparameters for subjective Bayesian analysis and
one often resorts to noninformative or objective priors. There are several
important reasons to consider objective priors. "Standard" noninformative
priors such as Jeffreys' prior, reference priors, or matching priors could
be difficult or impossible for the semiparametric components. Even when we
can derive these "standard" objective priors, posterior distributions are
usually computationally intractable because these priors often depend on
sample sizes. We propose a class of objective priors for full Bayesian
hierarchical linear mixed models and apply to space-time models. The models
are computationally feasible. Applications to economic time-dependent
data, psychological time response models, breast cancer survival models in
Iowa would be illustrated.
COFFEE: 3:45 p.m., 104 Snedecor Hall