Interval estimations of software reliability and optimal release time based on better bootstrap confidence intervals

We discuss a bootstrap method for estimating software reliability and cost-optimal software release time by using a discretized software reliability growth model. And then, we conduct interval estimations of them by five types of bootstrap confidence interval estimation methods including so-called better bootstrap confidence intervals, such as a bootstrap-t and a bias-corrected and the accelerated confidence interval (BCa) method. In our numerical examples, we confirm that our bootstrap approach yields simulation-based probability distributions of model parameters, software reliability assessment measures, and cost-optimal software release time without deriving these probability distributions analytically.