Bootstrapping empirical functions (Q914293)

Using a weak approximation approach the validity of the bootstrap for general empirical functions on the real line is proved. The usual empirical distribution function and the empirical quantile function are contained as special cases. The general results are then utilized to show the asymptotic validity of bootstrap confidence-band estimation of functions on the real line which are different from the underlying distribution or quantile function. Twelve examples of concrete functions are discussed, among them various reliability, concentration and moment-type functions. Moreover, it is demonstrated that even a result of \textit{K. B. Athreya} [ibid. 15, 724-731 (1987; Zbl 0628.62042), see also Mathematical statistics theory and applications, Proc. World Congr. Bernoulli Soc., Tashkent/USSR 1986, Vol. 2, 95-98 (1987; Zbl 0669.62023)] on the validity of bootstrapping the mean of a distribution with infinite variance from the domain of attraction of the normal distribution is covered by the theory presented in the paper.