Наукові вісті Національного технічного університету України "Київський політехнічний інститут" (Sep 2017)
Surface of Maximums of AR(2) Process Spectral Densities and its Application in Time Series Statistics
Background. In the problem on probabilities of large deviations of discrete time and sub-Gaussian AR(2) noise nonlinear regression model parameter least squares estimate a constant is determined that controls the rate of exponential convergence to zero of indicated probabilities. Objective. The aim of the paper is to find the surface of maximums of AR(2) process spectral densities in the domain of its stationarity in explicit form. Methods. The results were obtained on the use of methodology developed in the works by A. Sieders, K. Dzhaparidze (1987), A.V. Ivanov (1997, 2016) and standard Calculus methods. Results. A complex formula that describes a continuous surface of maximums of AR(2) process spectral densities assigned on the stationary triangle of the time series of this type is obtained. Conclusions. The obtained formula of surface of maximums of noise spectral densities gives an opportunity to realize for which values of AR(2) process characteristic polynomial coefficients it is possible to look for greater rate of convergence to zero of the probabilities of large deviations of the considered estimates.