Bayesian Modeling of Population Variability -- Practical Guidance and Pitfalls
With the advent of easy-to-use open-source software for Markov chain Monte Carlo (MCMC) simulation, hierarchical Bayesian analysis is gaining in popularity. This paper presents practical guidance for hierarchical Bayes analysis of typical problems in probabilistic safety assessment (PSA). The guidance is related to choosing parameterizations that accelerate convergence of the MCMC sampling and to illustrating the potential sensitivity of the results to the functional form chosen for the first-stage prior. This latter issue has significant ramifications because the mean of the average population variability curve (PVC) from hierarchical Bayes (or the mean of the point estimate distribution from empirical Bayes) can be very sensitive to this choice in cases where variability is large. Numerical examples are provided to illustrate the issues discussed.
- Research Organization:
- Idaho National Lab. (INL), Idaho Falls, ID (United States)
- Sponsoring Organization:
- USDOE
- DOE Contract Number:
- DE-AC07-99ID-13727
- OSTI ID:
- 933190
- Report Number(s):
- INL/CON-08-14208; TRN: US200814%%266
- Resource Relation:
- Conference: PSAM-9,Hong Kong,05/18/2008,05/23/2008
- Country of Publication:
- United States
- Language:
- English
Similar Records
Bayesian Inference in Probabilistic Risk Assessment -- The Current State of the Art
Bayesian network ensemble as a multivariate strategy to predict radiation pneumonitis risk