Assessing the quality of bootstrap samples and of the bootstrap estimates obtained with finite resampling
Journal
Statistics & Probability Letters
Date Issued
October 1, 2002
Author(s)
DOI
10.1016/S0167-7152(02)00196-7
Abstract
It is seen in simulations and confirmed theoretically that: (i) the loss in accuracy of the Monte Carlo approximation of the bootstrap estimate can be infinite, due to the additional uncertainty introduced by finite resampling, and (ii) the dimension of the data or the estimate of interest affect drastically the quality of the bootstrap samples and estimates. Based on the findings, directions are provided to improve the bootstrap methodology.