Asymptotic maximal deviation of M-smoothers (Q1117633)

Let \((X_ 1,Y_ 1),...,(X_ n,Y_ n)\) be i.i.d. rv's and let \(m(x)=E(Y\mid X=x)\) be the regression curve of \(Y\) on \(X\). An M-smoother \(m_ n(x)\) is a robust, nonlinear estimator of \(m(x)\), defined in analogy to robust M-estimators of location. In this paper the asymptotic maximal deviation \(\sup_{0\leq t\leq 1}| m_ n(t)-m(t)|\) is considered. The derived result allows the construction of a uniform confidence band for \(m(x)\).