No label defined (Q4518573)

Let \((X_{\Delta},Y_{\Delta})= \{(X_{\Delta}(t),Y_{\Delta}(t))\), \(t\in\mathbb R\}\), \(\Delta>0,\) be a family of two-dimensional zero-mean separable stationary Gaussian processes and let \(H=\{H(t), t\in\mathbb R\}\in L_2(\mathbb R)\) be a real-valued function. Suppose that the marginal spectral densities of the processes \(X_{\Delta}\) and \(Y_{\Delta}\) are \(f_{X_{\Delta}}(\lambda)=f_{\Delta}(\lambda)\) and \(f_{Y_{\Delta}}(\lambda)=|H^{\star}|^2f_{\Delta}(\lambda)\) and the cross spectral density is \(f_{{Y_{\Delta}}{Y_{\Delta}}}(\lambda)= H^{\star}f_{\Delta}(\lambda),\lambda\in\mathbb R.\) For a given \(\tau\in\mathbb R\) and \(\Delta>0\), and a given positive integer \(N\) let \(h=h(\Delta):=\pi/(2\Delta)\) and let the random variables \(\widehat{H}_{N,\Delta}(\tau)\) and \(\widehat{Z}_{N,\Delta}(\tau)\) be defined as \[ \widehat{H}_{N,\Delta}(\tau):={1\over{cN}}\sum_{n=1}^{N} Y_{\Delta}(nh)X_{\Delta}(nh-\tau),\qquad \widehat{Z}_{N,\Delta}(\tau):=\sqrt{Nh(\Delta)} \left[\widehat{H}_{N,\Delta}(\tau)-E\widehat{H}_{N,\Delta}(\tau)\right]. \] The paper deals with the asymptotic behaviour of the mean and the correlation function of the process \(\widehat{H}_{N,\Delta}(\tau)\) as \(\Delta\to\infty\) and \(N\to\infty.\) Asymptotic normality of finite-dimensional distributions of the processes \(\widehat{H}_{N,\Delta}(\tau)\) and \(\widehat{Z}_{N,\Delta}(\tau)\) is proved.