Confidence interval for the bootstrap P-value and sample size calculation of the bootstrap test

The use of the bootstrap test lends both elegance and simplicity to the analysis of complicated statistical problems. Such a numerical approach yields an estimated P-value as a binomial proportion. The purpose of this article is to recommend that confidence intervals for the bootstrap P-value be calculated routinely. In constructing these intervals, due consideration needs to be given to the P-values that may be close to or equal to zero. We consider three instructive examples and related methods for designing the bootstrap sample size.