A central limit theorem for nonuniform \(\varphi\)-mixing random fields with infinite variance (Q5933621)

Let \((\xi_{\mathbf k})\), \({\mathbf k}=(k_1,\dots,k_d)\in \mathbb{Z}^d\), be a random field. Denote by \(\mathcal B\) the class of Borel subsets of \([0,1]^d\). Consider the smoothed set-indexed partial sums processes \[ S_n(A)=\sum_{\overline 1\leq {\mathbf k}\leq\overline n}\frac{|A\cap C_{n,{\mathbf k}}|}{|C_{n,{\mathbf k}}|} (\xi_{\mathbf k}-E\xi_{\mathbf k}),\quad A\in {\mathcal B},\;n\in\mathbb{Z}_+, \] where \(C_{n,{\mathbf k}}=\prod^d_{i=1}((k_i-1)/n,k_n/n]\) and \(\overline{x} = (x,\dots,x)\). For certain nonuniform \(\varphi\)-mixing stationary random fields \((\xi_{\mathbf k})\) with infinite variance, the author shows that the processes \(Z_n(A) = S_n(A)/a_n\), for suitable normalizing constants \(a_n\) converge in the sense of finite-dimensional distributions to a Brownian motion on \(\mathcal B\). Examples based on Gibbs fields are given.