Gaussian approximation of mixing random fields (Q797219)

The author elaborates a natural two-parameter analogue of standard a.s. invariance principles for mixing sequences. To go into details, let \(\{\xi_{\nu}:\quad \nu \in N^ 2\}\) be a weakly stationary random field where \(N^ 2=\{\nu =(i,h):\quad i,k=1,2,...\}.\) Assume \(E\xi_{\nu}=0\), \(E| \xi_{\nu}|^{2+\delta}\leq C_ 1<\infty \quad(\nu \in N^ 2),\quad \rho(H_ 1,H_ 2)=\sup | P(AB)- P(A)P(B)| \leq C_ 2(\inf_{\mu \in H_ 1,\nu \in H_ 2}| \mu -\nu |)^{-\gamma},\) the supremum is taken over all \(A\in \sigma \{\xi_{\mu}:\quad \mu \in H_ 1\}\) and \(B\in \sigma \{\xi_{\nu}:\quad \nu \in H_ 2\}; H_ 1,H_ 2\) are disjoint sets in \(N^ 2\); \(0<\delta \leq 1\), \(\gamma \geq C_ 0/\delta\) and \(C_ 0,C_ 1,C_ 2\) are constants. Set \(S_ n=\sum_{\nu \leq n}\xi_{\nu}\) where the relation \(\nu \leq n\) is meant coordinatewise. The main result is the following. Let \(\{H(n): n\in N^ 2\}\) be a stationary Gaussian field with mean zero satisfying simple regularity conditions and \(\lim_{[n]\to \infty}ES^ 2_ n/EH(n)^ 2=1\) where \([n]=n_ 1n_ 2\). Then \((*)\quad S_ n- H(n)\ll [n]^{{1\over2}}(\log \log [n])^{-\lambda}\) a.s. holds for all \(n\in N^ 2\) where \(\lambda>0\). As usual, relation (*) is meant in the sense that the fields \(\{\xi_{\nu}\}\) and \(\{H(n)\}\) can be jointly defined on a suitable probability space without changing their distribution such that (*) holds. Various limit theorems connected with the law of iterated logarithm for the field \(\{\xi_{\nu}\}\) are deduced.