|
Activity Number:
|
77
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 2, 2009 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #303491 |
|
Title:
|
Neutral Noninformative and Informative Gamma, Beta, and Dirichlet Conjugate Priors
|
|
Author(s):
|
Jouni Kerman*+
|
|
Companies:
|
Novartis Pharma AG
|
|
Address:
|
Novartis Campus, Basel, International, 4002, Switzerland
|
|
Keywords:
|
Bayesian analysis ; prior distributions ; noninformative distributions ; conjugate priors ; estimation of proportions and rates
|
|
Abstract:
|
The commonly used conjugate noninformative Beta, Dirichlet, and Gamma distributions typically yield posterior tail probabilities that appear excessively skeptical or enthusiastic when compared to the observed proportion or rate. A default prior should not only feature little information but also be neutral in the sense that the observed rate be close to the posterior median regardless of the number of observations or the sample size. We introduce new conjugate noninformative priors Beta(1/3, 1/3) and Gamma(1/3, 0) having the special property that produces posterior medians approximately equal to the observed rate: they guarantee that the posterior probability that the parameter being less than the observed success rate is always between 0.49 and 0.51. We also introduce default Dirichlet priors with neutral marginals and discuss how to construct informative priors.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2009 program |
