Abstract:
The article describes numerical experiments aimed at calculating the spectral densities of two models of stationary stochastic processes on the basis of samples consisting of $n=2^{11}$ readings each. The effect on estimation quality of multiplying out the realization by some data window that smoothes the edges of the realization and of using the higher-order weighting functions proposed by the author in [Probl. Peredachi Inf., 1973, vol. 9, no. 4, pp. 42–48; 1980, vol. 16, no. 1, pp. 42–49; Comput. Appl. Math., no. 44, Vishcha Shkola, Kiev, 1981, pp. 32–40] is investigated.