PLACE:319 Snedecor Hall

SPEAKER:
Jaeyong Lee
National Institute of Statistical Sciences

TITLE:
Bayesian Bootstrap and its Application to Proportional Hazards Models

ABSTRACT:

Cox's proportional hazards model is a semiparametric model with a finite dimensional parameter of interest and an infinite dimensional nuisance parameter whose full Bayesian computation can be quite complicated. This is unfortunate because one has to spend most of one's time and effort in the complication arising from the nuisance parameter in which one has no or only mild interest.

The Bayesian bootstrap can be viewed as a data dependent likelihood approximation method such as the bootstrap and empirical likelihood.  Its extension to the proportional hazards model comes naturally along this line.

In this talk, I will try to show that the Bayesian bootstrap can be applied to proportional hazards models. It reduces the complexity of the computation significantly and gives a good approximation to the full Bayesian analysis. Furthermore, the approximate analysis can be interpreted as that with a noninformative prior.
 
 
 
 

COFFEE: 3:45 p.m., 104 Snedecor Hall