We consider nonparametric density estimation from the point of view of coverage probability. To take into account the problem of bias in bootstrapping nonparametric density kernel estimators, P. Hall [Statistics 22, No. 2, 215-232 (1991; Zbl 0809.62031); Ann. Stat. 20, No. 2, 675-694 (1992; Zbl 0748.62028)] showed that it is better to undersmooth the kernel density estimator than to estimate the
... [Show full abstract] bias and then to construct a bias-corrected bootstrap confidence interval. For a fourth-order kernel estimator, this leads to a two-sided bootstrap confidence interval with coverage probability error of size O(n -4/5 ) instead of O(n -12/17 ) in the latter case. We combine bias-correction with bootstrap without replacement as considered by D. Politis and J. P. Romano [Ann. Stat. 22, No. 4, 2031-2050 (1994; Zbl 0828.62044)] in order to get better rates.