The numerical bootstrap (Q2176627)

The authors introduce a new concept of a bootstrap method, denoted numerical bootstrap, intended to be an alternative to the commonly used \(m\)-out-of-\(n\) bootstrap mostly in cases where the standard bootstrap fails. After introducing the numerical bootstrap method, they compare it with the \(m\)-out-of-\(n\) bootstrap and subsampling and show a result on asymptotic convergence properties. The fourth section of the article is dedicated to the investigation of consistency of the numerical bootstrap for a class of M-estimators. The performance of the numerical bootstrap is proved by some Monte Carlo simulations. A list of used symbols, proofs and additional results are to be found in a Supplement to this article, under \url{doi:10.1214/19-AOS1812SUPP}.