skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Constrained Bayesian approach for testing TTBT compliance

Technical Report ·
OSTI ID:7242571

A constrained Bayesian approach is applied to the problem of statistically testing TTBT compliance. Our approach allows expert opinion to be used as prior information for the unknown parameters of the yield estimation problem. In addition, it is shown that existing seismic magnitude data may be used to furnish some information about the unknown parameters even though the associated yields are unknown. That is, since for equal slope parameters, msubb - msubLg does not depend on yield, one can estimate the mean and variance of this difference from data even though the yields are unknown. This information in included in the joint prior distribution in the form of constraints on the difference of the mean and variance parameters of the seismic magnitudes. The impact of such information on the test of compliance is investigated by comparing the power and F-numbers of two tests based on Bayesian approaches, with and without the constraints. The Bayesian results are also compared with those of two tests based on classical statistics. This formulation treats the intercepts and the covariance matrix of the seismic errors as unknown. To obtain closed form expressions for the power and F-numbers, however, the special case in which the intercepts are treated as random is considered, but the covariance matrix of the seismic errors is treated as known. The comparison shows that the probability of detecting a TTBT violation using the constrained Bayesian approach is always at least as great as when using the unconstrained approach.

Research Organization:
Mission Research Corp., Santa Barbara, CA (United States)
OSTI ID:
7242571
Report Number(s):
AD-A-253288/5/XAB; MRC-R-1367; CNN: F19628-90-C-0135
Resource Relation:
Other Information: Prepared in cooperation with Southern Methodist University, Dallas, TX
Country of Publication:
United States
Language:
English