Limit laws for extremes of dependent stationary Gaussian arrays (Q1933746)

Let \(\left(X_{n,k}^{(1)}, X_{n,k}^{(2)}\right)\), \(1\leq k\leq n\), \(n\in\mathbb N\), be a triangular array of bivariate Gaussian random vectors with zero-mean, unit variance and correlation \(\operatorname{corr}\left(X_{n,k}^{(1)}, X_{n,k}^{(2)}\right)=\rho(n)\), \(\operatorname{corr}\left(X_{n,k}^{(i)}, X_{n,l}^{(j)}\right)=\rho_{ij}(k,l,n)=\rho_{ij}(|k-l|,n)\). Under appropriate conditions on \(\rho_{ij}\), the limit distribution of \(\left(\max_{1\leq k \leq n}X_{n,k}^{(1)}, \max_{1\leq k \leq n}X_{n,k}^{(2)}\right)\), linearly standardized, is shown to be the bivariate Hüsler-Reiss distribution. This is well-known for independent vectors \(\left(X_{n,k}^{(1)}, X_{n,k}^{(2)}\right)\), \(1\leq k\leq n\). Under a strong dependence assumption, the limit of the maxima is a mixture of a bivariate Gaussian and Hüsler-Reiss distribution. The asymptotic independence of component-wise maxima and minima is preserved even in the case of weak dependence. Under the Berman condition for the components of the triangular array, an almost sure limit theorem is established.