Random Oper, and Stoch. Equ.f Vol. 3, No. 1, pp. 63-74 (1995)
© VSP 1995
On interpolation problem for vector-valued stochastic
sequences
M. P. MOKLYACHUK
Department of M echanics and M athematics, Kiev University, Kiev, Ukraine
Received for ROSE 15 December 1994
A bstract— The problem of the linear mean square optimal estimation is considered for the functional
3=0
of a Hilbert space valued stationary stochastic sequence £()), j £ Z from observations of the sequence
£ (J) + 17 (f) for j £ Z \ {0,1 .... , .V ]. Formulas are derived for computing the mean square
error and the spectral characteristic of the optimal estimate of the functional /4 V £. The minimax
spectral characteristic and the least favorable spect ral densities are found for various classes of spectral
densities.
1. IN T R O D U C T IO N
The classical Kolmogorov method of linear interpolation, extrapolation and filtering of
weakly stationary stochastic sequences [ 1], [2 ] may be employed under the condition that
spectral densities of stochastic sequences are known. In practice, however, the problem
of estimation of unknown values of stochastic sequence arises where the spectral density
is not known exactly. To solve the problem, the parametric or nonparametric estimate
of the unknown spectral density is found. Then the classical method is applied provided
that the estimate of the density is the true one. This procedure can result in a significant
increasing of the value of the error as Vastola and Poor have demonstrated with the
help of some examples [3]. For this reason it is necessary to search the estimate of
the unknown value of the stochastic sequence that has the least value of the error
for all densities from a certain class of spectral densities. Such an approach to the
problem of interpolation, extrapolation and filtering of stationary stochastic sequences
have been taken into consideration by many investigators [3] - [23]. A survey of results
in minimax (robust) methods of data processing can be found in the paper [9]. The
paper [ 10 ] is the first one where the minimax interpolation problem for the e - pollution
model is investigated. The relation of the minimax interpolation problem with the
problem of robust hypothesis testing is indicated in [8 ]. In papers [4] - [23] the minimax
interpolation, extrapolation and filtering problems are investigated with the help of the
convex optimization methods.
Translated by Yu. V. Kozachenko