|
Activity Number:
|
292
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #305740 |
|
Title:
|
Relationships between Frequentist and Bayesian Prediction Limits of the Poisson Process: Noninformative Priors
|
|
Author(s):
|
Valbona Bejleri*+
|
|
Companies:
|
University of District of Columbia
|
|
Address:
|
4200 Connecticut Ave., NW, Washington, DC, 20008,
|
|
Keywords:
|
prediction ; Bayesian ; Poisson ; prior ; non informative
|
|
Abstract:
|
Despite the different conceptual meaning of the frequentist and Bayesian approaches, relationships between the solutions these two approaches give to a specific prediction problem exist. Relationships between Bayesian prediction limits for Poisson distribution derived from noninformative priors and limits constructed in a frequentist way are presented. Steps on derivation of Bayesian limits are included for completeness. It is shown that the Bayesian lower limit derived from a modified Jeffreys' non informative prior is equal to the frequentist lower prediction limit and the frequency that the prediction will be wrong equals the Bayesian probability of wrong coverage. It is also shown that for any proper prior, there exist possible values of the observed outcome and the rate parameter such that the Bayesian lower limits do not coincide with the frequentist lower limits.
|
