On the smoothed bootstrap (Q1969149)

Let \(X_1,X_2,\dots, X_n\) be a sample of independent and identically distributed random variables. Denote by \(F\) the distribution function of \(X\). The authors consider the problem of estimating the parameter \(T(F)\) by a statistic \(T(\widehat F_n)\), where \(\widehat F_n(x)= \int^x_{-\infty} f_n(u) du\) and \(f_n(u)\) is the well-known kernel estimate of the unknown density function \(f(x)\). The main aim of this paper is to show that under suitable choice of the bandwidth parameter of \(f_n(x)\) confidence intervals based on the smoothed bootstrap have a smaller coverage error than those based on Efron's bootstrap.