Filtering Problem for Functionals of Stationary Processes with Missing Observations
2025
Read Filtering Problem for Functionals of Stationary Processes with Missing Observations (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-16908.txt.
Introduction
Mykhailo Moklyachuk and Maria Sidei study estimation with missing data in a 2025 paper. A stationary process is a random signal whose statistics stay constant. They want a weighted average of its unknown past values. Observations come from the signal plus noise, with some stretches missing.
When both spectra are known exactly, they give ready formulas. A spectrum here describes how the signal's energy spreads across frequencies. The formulas yield the best linear estimate and its error. When spectra are only known within allowed sets, they switch strategy. They find the worst allowed spectra and a robust estimate against them. That minimax estimate stays safe even under the least friendly conditions.
Catalog record
| YEAR | 2025 |
|---|---|
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 36 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-16908
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Filtering Problem for Functionals of Stationary Processes with Missing Observations Abstract Mathematical Research Press, London, UK Communications in Optimisation Theory, 2016, pp.1-18, Article ID 21 http://cot.mathres.org Mykhailo Moklyachuk ∗ , Maria Sidei , Department of Probability Theory, Statistics and Actuarial Mathematics, Taras Shevchenko National University of Kyiv, Kyiv 01601, Ukraine ∗ Corresponding Author: [email protected] Abstract. The problem of the mean-square optimal linear estimation of the functional A\xi=\ \int\limits_{R^{s}}a(t)\xi(-t)dt, which depends on the unknown values of stochastic stationary process \xi(t) from observations of the process \xi(t)+\eta(t) at points t\in\mathbb{R}^{-}\backslash S , S=\bigcup\limits_{l=1}^{s}[-M_{l}-N_{l},\,\ldots,\,-M_{l}], R^{s}=[0,\infty)\backslash S^{+}, S^{+}=\bigcup\limits_{l=1}^{s}[M_{l},\,\ldots,\,M_{l}+N_{l}] is considered. Formulas for calculating the mean-square error and the spectral characteristic of the optimal linear estimate of the functional are proposed under the condition of spectral certainty, where spectral densities of the processes \xi(t) and \eta(t) are exactly known. The minimax (robust) method of estimation is applied in the case where spectral densities are not known exactly, but sets of admissible spectral densities are given. Formulas that determine the least favorable spectral densities and the minimax spectral characteristics are proposed for some special sets of admissible spectral densities. Keywords: Stationary process, Mean square error, Minimax-robust estimate, Least favorable spectral density, Minimax spectral characteristic. Mathematics Subject Classification: Primary: 60G10, 60G25, 60G35, Secondary: 62M20, 93E10, 93E11 1 Introduction Effective methods of solution of the estimation problems (interpolation, extrapolation and filtering) for stationary stochastic sequences and processes were developed by A. N. Kolmogorov [ 13 ] , N. Wiener [ 34 ] , A. M. Yaglom [ 37 , 38 ] . An important contribution to the theory of estimation was made by H. Wold [ 35 , 36 ] , Yu. A. Rozanov [ 31 ] , E. J. Hannan [ 8 ] . The basic assumption of most of the developed methods of estimation of the unobserved values of stochastic processes is that the spectral densities of the considered stochastic processes are exactly known. However, in practice, these methods can not be applied since the complete information on the spectral densities is impossible in most cases. In order to solve the estimation problem one have to find parametric or nonparametric estimates of the unknown spectral densities. Then, under assumption that the selected or estimated densities are the true ones, one of the traditional estimation methods is applied. This procedure can result in significant increasing of the value of error as K. S. Vastola and H. V. Poor [ 33 ] have demonstrated with the help of some examples. To avoid this effect it is reasonable to search estimates which are optimal for all densities from a certain class of admissible spectral densities. These estimates are called minimax since they minimize the maximum value of the error. This method was first proposed in the paper by Ulf Grenander [ 7 ] where this approach to extrapolation problem for stationary processes was applied. Various models of spectral uncertainty and minimax-robust methods of data processing can be found in the survey paper by S. A. Kassam and H. V. Poor [ 12 ] . In their papers J. Franke [ 4 ] , J. Franke and H. V. Poor [ 5 ] investigated the minimax extrapolation and filtering problems for stationary sequences with the help of convex optimization methods. This approach makes it possible to find equations that determine the least favorable spectral densities for some classes of admissible densities. Papers by M. Moklyachuk [ 17 ] – [ 20 ] contain results of investigation of the problem of optimal estimation of the functionals which depend on the unknown values of stationary sequences and processes. M. Moklyachuk and A. Masyutka developed the minimax technique of estimation for vector-valued stationary stochastic processes in papers [ 22 ] – [ 24 ] . Methods of solution of the interpolation, extrapolation and filtering problems for periodically correlated stochastic processes were developed by M. Moklyachuk and I. Golichenko [ 21 ] . Estimation problems for functionals which depend on the unknown values of stochastic processes with stationary increments were investigated by M. Luz and M. Moklyachuk [ 14 ] – [ 16 ] . The interpolation problem for stationary sequence with missing values was investigated by M. Moklyachuk and M. Sidei [ 25 , 26 ] . Prediction problem of stationary processes with missing observations was investigated in papers by P. Bondon [ 1 , 2 ] , Y. Kasahara, M. Pourahmadi and A. Inoue [ 11 , 28 ] , R. Cheng, A. G. Miamee, M. Pourahmadi [ 3 ] . The interpolation problem for stationary sequences was considered in the paper of H. Salehi [ 32 ] . In this article we deal with the problem of the mean-square optimal linear estimation of the functional A\xi=\int\limits_{R^{s}}a(t)\xi(-t)dt, which depends on the unknown values of a stationary stochastic process \xi(t) from observations of the process \xi(t)+\eta(t) at points t\in\mathbb{R}^{-}\backslash S , S=\bigcup\limits_{l=1}^{s}[-M_{l}-N_{l},\ldots,-M_{l}], R^{s}=[0,\infty)\backslash S^{+}, S^{+}=\bigcup\limits_{l=1}^{s}[M_{l},\,\ldots,\,M_{l}+N_{l}], M_{l}=\sum\limits_{k=0}^{l}(N_{k}+K_{k}), N_{0}=0 , K_{0}=0 . The case of spectral certainty as well as the case of spectral uncertainty are considered. Formulas for calculating the spectral characteristic and the mean-square error of the optimal linear estimate of the functional are derived under the condition that spectral densities of the processes are exactly known. In the case of spectral uncertainty, where the spectral densities are not exactly known while a set of admissible spectral densities is given, the minimax approach is applied. Formulas that determine the least favorable spectral densities and the minimax-robust spectral characteristics of the optimal estimates of the functional are proposed for some specific classes of admissible spectral densities. 2 Hilbert space projection method of filtering Consider a stationary stochastic process \{\xi(t),t\in\mathbb{R}\} with absolutely continuous spectral measure F(d\lambda) and spectral density f(\lambda) . Consider another stationary stochastic process \{\eta(t),t\in\mathbb{R}\} , uncorrelated with the process \{\xi(t),t\in\mathbb{R}\} , with absolutely continuous spectral measure G(d\lambda) and spectral density g(\lambda) . Without loss of generality, we suppose that introduced processes have zero mean values E\xi(t)=0 , E\eta(t)=0 . Assume that the spectral densities f(\lambda) and g(\lambda) satisfy the minimality condition \int\limits_{-\infty}^{\infty}\frac{\left|\gamma(\lambda)\right|^{2}}{f(\lambda)+g(\lambda)}d\lambda<\infty, (1) where \gamma(\lambda)=\int\limits_{0}^{\infty}\alpha(t)e^{it\lambda}dt is a nontrivial function of exponential type. This condition guarantees that the mean-square errors of estimates of the functionals are nonzero. [ 31 ] . Stationary processes \xi(t) and \eta(t) admit the spectral decomposition [ 10 ] \xi(t)=\int\limits_{-\infty}^{\infty}e^{it\lambda}Z_{\xi}(d\lambda),\hskip 28.45274pt\eta(t)=\int\limits_{-\infty}^{\infty}e^{it\lambda}Z_{\eta}(d\lambda), (2) where Z_{\xi}(d\lambda) and Z_{\eta}(d\lambda) are orthogonal stochastic measures defined on [-\pi,\pi) that correspond to the spectral measures F(d\lambda) and G(d\lambda) , such that the following relations hold true EZ_{\xi}(\Delta_{1})\overline{Z_{\xi}(\Delta_{2})}=F(\Delta_{1}\cap\Delta_{2})=\,\frac{1}{2\pi}\int_{\Delta_{1}\cap\Delta_{2}}f(\lambda)d\lambda, EZ_{\eta}(\Delta_{1})\overline{Z_{\eta}}(\Delta_{2})=G(\Delta_{1}\cap\Delta_{2})=\,\frac{1}{2\pi}\int_{\Delta_{1}\cap\Delta_{2}}g(\lambda)d\lambda. The main purpose of the article is to find the mean-square optimal linear estimate of the functional A\xi=\int\limits_{R^{s}}a(t)\xi(-t)dt, which depends on the unknown values of the process \xi(t) , based on the observed values of the process \xi(t)+\eta(t) at points t\in\mathbb{R}^{-}\backslash S , where S=\bigcup\limits_{l=1}^{s}[-M_{l}-N_{l},\ldots,-M_{l}] , R^{s}=[0,\infty)\backslash S^{+}, S^{+}=\bigcup\limits_{l=1}^{s}[M_{l},\,\ldots,\,M_{l}+N_{l}] . Let the function a(t) which determines the functional A\xi satisfy the conditions \int\limits_{R^{s}}\left|a(t)\right|dt<\infty,\quad\int\limits_{R^{s}}t\left|a(t)\right|^{2}dt<\infty. (3) Due to the spectral decomposition ( 2 ) of the process \xi(t) , the functional A\xi can be represented in the form A\xi=\int\limits_{-\infty}^{\infty}A(e^{i\lambda})Z_{\xi}(d\lambda),\quad A(e^{i\lambda})=\int\limits_{R^{s}}a(t)e^{-it\lambda}dt. Consider the Hilbert space H=L_{2}(\Omega,\mathcal{F},P) generated by random variables \xi with zero mathematical expectation, E\xi=0 , finite variation, E|\xi|^{2}<\infty , and inner product (\xi,\eta)=E\xi\overline{\eta} . Denote by H^{s}(\xi+\eta) the closed linear subspace generated by elements \{\xi(t)+\eta(t):t\in\mathbb{R}^{-}\backslash S\} in the