|
Activity Number:
|
432
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 5, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
IMS
|
| Abstract - #303493 |
|
Title:
|
A Hierarchical Bayesian Model to Merge Prior Opinions
|
|
Author(s):
|
Maria J. Rufo Bazaga*+ and Jacinto Martin Jimenez and Carlos J. Perez Sanchez
|
|
Companies:
|
University of Extremadura and University of Extremadura and University of Extremadura
|
|
Address:
|
Avda. de la Universidad S/N, Caceres, 10071, Spain
|
|
Keywords:
|
Bayesian analysis ; conjugate prior distributions ; mixture of prior distributions ; Kullback-Leibler divergence
|
|
Abstract:
|
This work provides a general Bayesian approach for the analysis of quadratic natural exponential families in the presence of several sources of prior information, such as experts. Each prior belief is elicited as a conjugate prior distribution and a mixture model is considered to represent a consensus of several experts. A hyperprior distribution on the weights is given. Then, a general method based on the expected Kullback-Leibler divergence to obtain the hyperparameters is proposed. This procedure leads to analytical solutions. Besides, a direct implementation for all families is possible. Finally, the expected discrepancies between the combined posterior distribution over the quantity of interest and each expert's prior distribution are analyzed. A computationally low-cost approach is proposed to estimate them.
|
