scientific article; zbMATH DE number 193533

From MaRDI portal
Publication:4040333





Let P(\(\cdot | t)\), \(t\in [0,1]\), be the (unknown, but determined) conditional distributions. \(X_ 1(t_ 1),...,X_ n(t_ n)\) are observed for a given design \(t^{(n)}=(t_ 1,...,t_ n)\). The value \(X_ i=X_ i(t_ i)\) can be written in the form \[ X_ i=R(t_ i)+\xi_ i(t_ i),\quad R(t)=E\{X_ i(t_ i)| t_ i=t\}, \] where R(t) is the regression function. The problem is to estimate the value F(R), F being a known functional defined on \(L_ 2[0,1]\), by means of the known values \(t_ 1,X_ 1,...,t_ n,X_ n.\) The author obtains asymptotical minimax lower bounds of mean square risk for estimators of a differentiable functional on regression with the possibility of experimental design. The sample design and the estimator, for which this bound is asymptotically tight, are constructed in the case of the known conditional variance of the ``noise, if some conditions of smoothness for the functional and the unknown regression function are satisfied.











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 Q4040333)