Weak convergence of the U-statistic and weak invariance of the one-sample rank order statistic for Markov processes and ARMA models (Q908623): Difference between revisions

Let \(X_{ni}\), \(1\leq i\leq n\), be possibly dependent, nonidentically distributed continuous random variables. For the random vector \(\underset \tilde{} X_ n=(X_ 1,...,X_ n)\) let \(U_ n\) denote a U- statistic and \(S_{n,m}\) be a rank order statistic of the form \[ \sum^{m}_{i=1}h(i/n)s(X_{ni})J(R_{n,m,i}/(m+1)),\quad 1\leq m\leq n, \] where h is continuous, \(s(x)=sign(x)\), J is a score function and \[ R_{n,m,i}=\sum^{m}_{j=1}I_{[| X_{nj}| \leq | X_{ni}|]},\quad 1\leq i\leq m\leq n. \] The authors [ibid. 30, 181-204 (1989)] have shown that \(U_ n\) is asymptotically normal, and \(S_{n,m}\) is asymptotically a function of a Brownian motion process, under some regularity conditions. Here it is shown that these conditions are satisfied when \(X_{ni}'s\) form either an aperiodic, Harris recurrent and geometrically ergodic or an aperiodic and Doeblin recurrent Markov chain, or when they form an ARMA process.