Jackknifing Incomplete U-statistics

Abstract:

      In this paper, we propose jackknife only for a particular class of non-symmetric statistics, which is incomplete U-statistics of significance in practice. We find that if we deal with incomplele U-statistics using a slightly modified jackknife, then Tu-key's conjecture will be true, and the consistent estimator of asymptotic variance is given. Thus we obtain an approximate confidence interval for reguluar function g (θ).