Some results on the influence of extremes on the bootstrap (Q1209197)

For a random sample of size \(n\), the authors consider the problems of bootstrapping consistently (i) the sample maximum, (ii) intermediate trimmed means where the trimming proportions are \(o(n)\), and (iii) the sample mean where the population is in the domain of attraction of a stable law. While doing so, they explore the influence of the extreme order statistics on bootstrapping (ii) and (iii), and establish some interconnections between the normalizing constants. The proofs depend heavily on empirical process methodology developed by the authors and their coworkers.