Pointwise and uniform convergence with probability 1 of nonparametric regression estimators (Q806855)

The authors study the pointwise and uniform convergence with probability one for nonparametric regression estimators constructed with the help of a convex function \(\Phi\). When \(\Phi (u)=u^ 2\) one gets the Nadaraya- Watson estimator, and for any other function one gets a nonlinear nonparametric M-estimator. It is shown that if \(\delta_ n\) is the estimation window, under certain regularity conditions for the functions involved one has pointwise convergence w.p.1 for \[ \delta_ n\to 0\text{ and } n \delta_ n^ d/\log \log n\to +\infty. \] If \(B\subset {\mathbb{R}}^ d\), under certain additional hypothesis and for \[ \delta_ n\to 0,\quad n \delta_ n^{d+1/p}/\log n\to +\infty \] there is uniform convergence on B w.p.1.