A pivotal function approach to estimation and prediction for a model of software reliability
Kelly J. Dotson · 1987
This paper establishes the utility of pivotal functions for assessing the reliability of software that has undergone a testing and development process. Confidence limits for reliability and prediction limits for the time to next failure are derived based on Moranda's geometric deutrophication model. The effect of departures from the assumed exponentially distributed interfailure times in this model is investigated by simulating Pareto, Weibull, and gamma distributed interfailure times. As a result of the simulation, the distributional form of the interfailure times is shown to significantly affect the confidence level of the prediction limits and suggests restricted use of the Moranda model.