The Use of Agent Judgements in Reliability Analysis

Authors: 
J. Crosse, M. West
Duke University, University of Warwick

Nov 30 1989

Bayesian models are described for problems in which an analyst is to assess a survival function for a system and receives relevant information from an individual (expert or agent), such information representing the agent's opinions about quantities related to the survival function. Analysis shows how, based only on the specification of a prior survival function and expectations for the agent's statement, the analyst may fully use such agent information. Types of agent information admitted by the approach here include collections of probability from single agent, complete agent probability distributions or densities, and the same from two or more agents. Results are especially useful when considering reliability of new systems for which no objective data is available.

Keywords: 

Bayesian updating, Combination of forecasts, expert/agent opinion, systems reliability analysis

Manuscript: 

PDF icon 1990-08.pdf