scientific article; zbMATH DE number 4009605

From MaRDI portal
Publication:3759763





For wide-sense stationary discrete-time random sequences \(\{X_ t\}\) with mean 0 and square-integrable spectral density f(\(\cdot)\) estimated by \(\hat f{}_ N(\cdot,X_ 1,...,X_ N)\), the authors define maximal- risks \[ R_ N\equiv R_ N(\hat f,{\mathcal X})\equiv \sup_{X\in {\mathcal X}}E_ X\| \hat f-f\|^ 2_ 2,\quad r_ n\equiv r_ N(\hat f,{\mathcal X})\equiv \sup_{X\in {\mathcal X}}E_ X\| \hat f-f\|^ 2_ 2/\| f\|^ 2_ 2. \] Here \({\mathcal X}\) is a specified collection of wide-sense stationary random sequences with mean 0 and square-integrable spectral density whose covariances and fourth semi-invariants satisfy some further summability conditions. The purpose of the paper is to announce and to give examples of results telling the top-order asymptotic behavior for large N of the infima \[ \Delta^ 2_ N = \inf_{\hat f}R_ N(\hat f,{\mathcal X}),\quad \delta^ 2_ N\quad =\inf_{\hat f}r_ N(\hat f,{\mathcal X}) \] over all possible estimators \(\hat f(\cdot)= \hat f_ N(\cdot,X_ 1,...,X_ N)\).











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3759763)