Estimates of the correlation function of a Gaussian stationary process when its values are known in a finite set of points (Q6600510)

The authors deal with the problem of estimation of the correlation function of a Gaussian stationary random process \(X(t)\) with unknown mean value based on observations of its values at points \(t_k=kT/n,k=0,1,\dots,n\) and at points \(t_k+\tau\). We cannot use correlogram as an estimator since the correlogram is a biased estimation of the correlation function in this case. The authors use statistics \(\widehat{\rho}_{T,n}(\tau)=\sum_{k=0}^{n-1}\left(X\left({kT}/{n}+\tau\right)-\hat{a}_{\tau}\right)\left(X\left({kT}/{n}\right)-\hat{a}_0\right)/n,\) \(\hat{a}_{\tau}=\sum_{k=0}^{n-1}X\left({kT}/{n}+\tau\right)/n\) to estimate the correlation function \(\rho(\tau)=E(X(t+\tau)-m)(X(t)-m)\). Theory of square-Gaussian random processes (see, for example, \textit{Y. V. Kozachenko} and \textit{O. M. Moklyachuk}, Theory Stoch. Process. 6(22), No. 3--4, 98--121 (2000; Zbl 0977.60042); Extremes 2, No. 3, 269--293 (1999; Zbl 0961.60043)) is used to obtain estimates of the deviation of the proposed estimate from the correlation function of a Gaussian stationary random process with an unknown mean in \(L_p\)-metric. A criterion for testing hypothesis about the correlation function of such a random process is formulated based on the obtained estimates. The results described in this paper are based on the results obtained by \textit{Y. Kozachenko} and \textit{V. Troshki}, Mod. Stoch., Theory Appl. 1, No. 2, 139--149 (2014; Zbl 1349.62384); Commun. Stat., Theory Methods 47, No. 18, 4556--4567 (2018; Zbl 1508.62208).