Bayesian Decision Models for System Engineering

Abstract
This paper shows how modern developments in statistical decision theory can be applied to a typical systems engineering problem. The problem is how to design an experiment to evaluate a reliability parameter for a device and then make a decision about whether to accept a contract for the development and maintenance of a system of these devices. We introduce the concept of subjective probability distribution to permit encoding prior knowledge about the uncertainty in the process. The expected value of clairvoyance is computed as an upper bound to the value of any experimental program. The structure of decision trees serves as a means for establishing the optimum size and type of experimentation and for acting on the basis of experimental results. The subjective probability approach to decision processes allows us to consider and solve problems that previously we could not even formulate.