Bivariate symmetric statistics of long-range dependent observations (Q1176759): Difference between revisions

For a stationary, mean-zero Gaussian sequence \(\{X_ i\}_ 1^ \infty\) with long-range dependence, i.e. with covariance \(r(k)=EX_ 1X_{k+1}\), satisfying \(r(0)=1\); \(r(k)=k^{-D}L(k)\), \(0<D<1\), where \(L\) is slowly varying at infinity and positive for all large \(k\), and \(Y_ j=G(X_ j)\), the \(U\)-statistic \[ U_ N(h)=\sum_{{1\leq j_ 1,j_ 2\leq N}\atop{j_ 1\neq j_ 2}}h(Y_{j_ 1},Y_{j_ 2}) \] and the von Mises statistic \[ V_ N(h)=\sum_{j_ 1=1}^ N\sum_{j_ 2=1}^ Nh(Y_{j_ 1},Y_{j_ 2}) \] are considered. The asymptotic distributions of these statistics are considered under regularity conditions on \(h\) relaxed to ``locally bounded total variation''.