Asymptotics for linear random fields (Q5933609): Difference between revisions

Consider a multiparameter linear random field \(u(t)= \sum_i a(i)\xi(t- i)\), \(i,t\in Z^p_+\), where \(\xi(t)\) are i.i.d. zero-mean random variables with finite \(q\)th moment, \(q> 2p\), and \(Z\) is the set of integers. Under certain summation condition on \(a(i)\) two results are proved: a functional central limit theorem (\(n^{-p/2} \sum^{[nr]}_{t= 1} u(t)\), \(r\in [0,1]^p\), converges weakly to a multiparameter Brownian motion on \([0,1]^p)\) and strong approximation theorem (\(|\sum^{[nr]}_{t=1} (u(t)- \gamma(t))|= O(n^{(p- \delta)/2}(\log n)^{1/2})\) a.s. for all \(r\in [0,1]^p\), where \(\gamma(t)\) is i.i.d. zero-mean Gaussian random field). The main idea of the proof is a decomposition of partial sums of \(u(t)\) to a sum of independent innovations and a small remainder. This extends to \(p\geq 2\) a technique developed by \textit{P. C. B. Phillips} and \textit{V. Solo} [Ann. Stat. 20, No. 2, 971-1001 (1992; Zbl 0759.60021)].