Asymptotic behavior of \(L\)-statistics for a large class of time series (Q1335372)

Let \(\{X_ i\}\) be an \(m(n)\)-decomposable process of real valued random variables with a common distribution function \(F\). Thus, for each \(n\in\mathbb{N}\) we have a decomposition \(X_ i= X_{i,m(n)}+ \overline{X}_{i,m(n)}\) for some \(m(n)\in \{0,1, \dots,n\}\), where the \(X_{i,m(n)}\) have a common distribution function \(F_{m(n)}\), and where the \(X_{i,m(n)}\) are \(m(n)\)-dependent and \[ \max_{1\leq i\leq n} P(| \overline{X}_{i,m(n)}|\geq \varepsilon(n))\leq \delta(n) \qquad \text{for some } \varepsilon(n), \quad \delta(n)\to 0 \text{ as } n\to \infty. \] Suppose \(F\) is continuously differentiable on \(R\) with derivative \(f\) satisfying \(\| f\|_ \infty <\infty\) and \(\{x\in R\): \(f(x)>0\}\) convex. Let \(X_{1:n}, \dots, X_{n:n}\) be order statistics of \(X_ 1,\dots, X_ n\). Let \(\Psi: (0,1)\to R\) and \(J: (0,1)\to R\) be given functions and consider linear combinations of functions of order statistics (L-statistics) of the type \[ T_ n= n^{- 1} \sum^ n_{i=1} c_{n,i} \Psi(X_{i:n})= n^{-1} \sum^ n_{i=1} c_{n,i} \Psi_ F(\xi_{i:n}) \] where the \(c_{n,i}\) are constants defined by \(J\), \(\xi_ i= F(X_ i)\) and \(\Psi_ F= \Psi(F^{-1})\). The authors derive the asymptotic normality of \(T_ n\) under some conditions on \(J\) and \(\Psi\), using a new method to specify the dependence structure. The result is valid for \(L\)-statistics of a wide class of time series.