On the prediction of vector-valued random fields and the spectral distribution of their evanescent component (Q2507750)

Let \((X_{m,n})_{(m,n)\in \mathbb{Z}^2}\) be a \(C^p\)-valued wide sense stationary process. The prediction theory of such processes according to different total orders on \( \mathbb{Z}^2\) is studied. The goal in prediction theory is, given the ``past'' of the process, to approximate the process by its past and to evaluate the prediction error committed or the ``innovation'' of the process. In the case of the processes indexed by \( \mathbb{Z}^2,\) one matter is that several natural definitions of ``past'' may be given. The main result is obtained in the case where the past is defined by the ``usual lexicographic order''. In particular, a spectral condition (generalizing the condition of ``full rank'' for non-determinism) under which the evanescent component is non-trivial is given and then the spectral distribution of the evanescent component is identified. ``Rational'' orders different from the usual lexicographic are investigated. It is shown, in particular, that for two different rational orders the resulting evanescent components are mutually orthogonal. The proposed method may be applied for processes indexed by \( \mathbb{Z}^l\) \((l\geq2)\) with the corresponding notational complications. An example of a purely evanescent process, with respect to some ``non-rational'' order, whose spectral distribution is not absolutely continuous with respect to the Lebesgue measure, is given.