Distributed statistical inference for massive data

This paper considers distributed statistical inference for general symmetric statistics in the context of massive data with efficient computation. Estimation efficiency and asymptotic distributions are provided, which reveal different results between nondegenerate degenerate cases, show number subsets plays an important role. Two bootstrap methods proposed analyzed to approximation underlying distribution improved computation over existing methods. The accuracy distributional by studied theoretically. One methods, pseudo-distributed bootstrap, is particularly attractive if datasets large as it directly resamples subset-based statistics, assumes less stringent conditions its performance can be studentization.