Bandwidth selection for the smoothed bootstrap percentile method. (Q5941109)

Some applications of the bootstrap involve smoothing the estimated distribution that is resampled, a method known as the smoothed bootstrap. Recently, the effect of resampling a kernel smoothed distribution was evaluated through expansions for the coverage of bootstrap percentile confidence intervals. It was shown that under a smooth function model, a bandwidth of order \(n^{-1/4}\) can accomplish a first-order correction for the one-sided percentile method. This paper proposes and investigates the performance of some data-based methods for selecting this bandwidth. The methods are studied through asymptotic comparisons and the resulting intervals are compared in the finite sample case to other well-known bootstrap confidence interval methods through a simulation study.