BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:The Blended Paradigm:  A Bayesian Approach to Handling Outliers an
 d Misspecified Models  - Prof. Steven MacEachern (Ohio State University)
DTSTART:20141215T100000Z
DTEND:20141215T110000Z
UID:TALK56674@talks.cam.ac.uk
CONTACT:Zoubin Ghahramani
DESCRIPTION:Bayesian methods have proven themselves to be enormously succe
 ssful across a wide range of scientific problems\, with analyses ranging f
 rom the simple one-sample problem to complicated hierarchical models. They
  have many well-documented advantages over competing methods. However\, Ba
 yesian methods run into difficulties for two major and prevalent classes o
 f problems—handling data sets with outliers and dealing with model missp
 ecification. In both cases\, standard Bayesian analyses fall prey to the h
 ubris that is an integral part of the Bayesian paradigm. The large sample 
 behavior of the analysis is driven by the likelihood. We propose the use o
 f restricted likelihood as a single solution to both of these problems. Wh
 en working with restricted likelihood\, we summarize the data\, x\, throug
 h a set of (insufficient) statistics T(x) and update our prior distributio
 n with the likelihood of T(x) rather than the likelihood of x. By choice o
 f T(x)\, we retain the main benefits of Bayesian methods while reducing th
 e sensitivity of the analysis to selected features of the data. The talk w
 ill motivate the blended paradigm\, discuss properties of the method and c
 hoice of T(x)\, cover the main computational strategies for its implementa
 tion\, and illustrate its benefits.  This is joint work with Yoonkyung Lee
  and John Lewis.
LOCATION:Engineering Department\, CBL Room BE-438.
END:VEVENT
END:VCALENDAR