Moment Consistency of the Exchangeably Weighted Bootstrap for Semiparametric M-estimation

摘要: The bootstrap variance estimate is widely used in semiparametric inferences. However, its theoretical validity a well-known open problem. In this paper, we provide first study on the moment estimates models. Specifically, establish consistency of Euclidean parameter, which immediately implies t-type confidence set. It worth pointing out that only additional cost to achieve contrast with distribution simply strengthen L1 maximal inequality condition required latter Lp for p≥1. general multiplier developed paper also independent interest. These conclusions hold methods exchangeable weights, example, non-parametric and Bayesian bootstrap. Our theory illustrated celebrated Cox regression model.