Exploring spectral density estimation for spatial linear process with mixing innovations (Q2302708)

Denote by \((X_t,t\in\mathbb{Z}^N)\) a stationary spatial linear process defined on a probability space \((\Omega,\mathcal{A},{P})\), of unknown spectral density \(f_X\), \(X_t=\Sigma_sa_sZ_{t-s}\), \(t\in\mathbb{Z}\), \(Z_t\) being a strictly stationary process with an unknown spectral density \(f_Z\). Denote by \(I_{n,X}\) the spatial periodogram of \(X_t\). The aim of the paper is to investigate asymptotic properties of spectral density estimates \(\hat{f}_X\) of \(X_t\) using observations on \(I_n\). Under some quite special assumptions, asymptotic properties of the spatial periodograms of \((X_t)\) and \((Z_t)\) are proved first. Then, one establishes a relation between the periodograms of \((X_t)\) and \((Z_t)\) and one proves a result concerning the asymptotic properties of the periodogram of \((X_t)\). By means of introducing a linear filter, a smoother periodogram estimator is obtained and its convergence properties are studied. Results on numerical experiments are shown in the last section of the paper.