No label defined (Q4939449)

Let \(X_1,X_2,\dots, X_n\) be i.i.d. random variables. Let \(S_n= S(X_1,X_2,\dots, X_n)\) be a symmetric statistic related to a parameter \(\theta_n\). Define a jackknife variance estimator \(\widehat\sigma^2_n\) by \[ \widehat\sigma^2_n= (n-1) \sum^n_{i=1} (S^{(i)}_n- S_n)^2, \] where \(S^{(i)}_n\) denotes the symmetric statistic computed from a sample of \(n-1\) points with \(X_i\) left out. The author establishes the following: \[ \sup_x|P\{\widehat\sigma_n \sqrt n(S_n- \theta_n)\leq x)- \Phi(x)|= O(n^{-1/2}), \] where \(\Phi(x)\) denotes the standard normal distribution and \[ \sup_x|P\{\widehat\sigma_n \sqrt n(S_n- \theta_n)\leq x)-H_n(x)|= o(n^{-1/2}), \] where \(H_n(x)\) denotes the Edgeworth expansion.