|
Activity Number:
|
241
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #306933 |
|
Title:
|
New Priors for Complex System Reliabilities and Trends
|
|
Author(s):
|
Earl Lawrence*+ and Scott Vander Wiel
|
|
Companies:
|
Los Alamos National Laboratory and Los Alamos National Laboratory
|
|
Address:
|
Statistical Sciences Group, D1, Los Alamos, NM, 87545,
|
|
Keywords:
|
prior ; component ; system ; reliability
|
|
Abstract:
|
When modeling reliability degradation in complex systems it is tempting to use diffuse priors on parameters of the system's components and subcomponents. This can lead to a nonsensical system prior and have a dramatic effect on the posterior when data are sparse. We explore negative log-gamma component (NLG) priors that induce a more appropriate a priori distribution of system reliability and its evolution over time. NLG distributions are similar to the beta, but closed under products. This closure makes NLGs ideal for modeling complex series systems in which the system-level reliability is a product of the component-level reliabilities. The NLG model is easily extended to include a hierarchical prior that anticipates situations where only a few of the components will contribute substantially to the level and trend of system reliability over time.
|
