A.J. Canty, A.C. Davison, D.V. Hinkley, and V. Véntura
Bootstrap diagnostics are used to assess reliability of bootstrap calculations, and may suggest useful modified calculations when these are possible. Concern focuses on susceptibility to peculiarities in data, incorrectness of resampling model, incorrect use of resampling simulation output, and inherent inaccuracy of the bootstrap approach. The last involves issues such as inconsistency of bootstrap method, order of correctness of a consistent bootstrap method, and approximate pivotality. In this paper we attempt to address these problems, to provide workable diagnostic methods, and where necessary to provide fast and simple ways to effect the necessary computations.
Keywords: Blocking; Bootstrap recycling; Breakdown; Count data; Extreme values; Importance resampling; Inconsistency; Jackknife-after-bootstrap; Linear approximation; Long-tailed distributions; Nested bootstrap; Outlier; Pareto distribution; Periodogram; Pivot; Resampling model; Robust estimate; Shrinkage estimator; Smoothers; Stein estimator; Stratification; Superefficiency; Tiling; Time series; Variance function; Wavelet regression.