Citation

Edwards G. Reliab. Eng. 1984; 9(1): 1-17.

Copyright

(Copyright © 1984, Elsevier Publishing)

DOI

10.1016/0143-8174(84)90002-7

PMID

unavailable

Abstract

This paper investigates a Bayesian procedure for drawing inferences from random data. The procedure is particularly relevant in any reliability application where the available data set is small. This is because it allows the modelling uncertainty, concerning the choice of distribution type, and the statistical uncertainty, concerning the distribution parameter estimates, to be considered alongside the underlying physical uncertainty when drawing the inferences. The calculations are carried out by considering a series of possible distribution types and a series of possible sets of parameters (parameter vectors) within each distribution. An 'expected' inference is then taken over all distribution/parameter vector combinations.

The procedure is illustrated via the analysis of a simple structural reliability problem for which exact answers are known.