Our objective is to review the results on bootstrap theory for U-statistics. The focus is on consistency properties of bootstrap estimators for the distribution function of properly standardized (or studentized) U-statistics. Simple examples are used to explain the reason for bootstrap success or bootstrap failure. General resampling schemes are considered. A survey of material on bootstrapping U-statistics not discussed in the main sections of this paper is included as final section.
机构:
St Petersburg Univ Means Commun, Dept Appl Math, St Petersburg 190031, RussiaSt Petersburg Univ Means Commun, Dept Appl Math, St Petersburg 190031, Russia