Estimates of Reliability During the Test and Evaluation Stage
Some Methodological Observations
Within the acquisition process, reliability testing is one of the methods the military has used to maintain control of the reliability characteristics of a system being developed. Reliability testing has multiple objectives, including (1) determining compliance with contractual requirements; (2) identifying deficiencies in the system; (3) providing reliability measures for use in operation planning; and (4) measuring the readiness of the system for production and operational use. Applying the statistical techniques generally used, this paper illustrates the statistical risk facing both the military purchaser and the contractor from testing less, and conversely, the gains in statistical confidence from testing more. It considers two basically different kinds of reliability: (1) binomial processes, where "success" or "failure" are defined and testing is required to estimate the probability of their occurrence and (2) time-to-failure reliability, where "failure" is defined and the times when failures occur are recorded. The author concludes that both these approaches are likely to require some increase in the size of test programs.