On the \(L_ p\) norm for some approximation operators (Q1065301): Difference between revisions

Let \(X_ 1,X_ 2,..\). be random variables with mean x and variance \(\sigma^ 2(x)\) where x is a continuous parameter with values in \(I\subseteq {\mathbb{R}}=(-\infty,\infty)\). For any continuous bounded function f on \({\mathbb{R}}\) define the Feller operator [cf. the author, Acta Math. Hung. 35, 193-203 (1980; Zbl 0404.41012)] by \(L_ n(f,x)=Ef(S_ n/n)=\int^{\infty}_{-\infty}f(t/n)dF_{n,x}(t),\) where \(F_{n,x}(t)\) is the distribution function of \(S_ n=X_ 1+...+X_ n\). Also, define \[ D^ p_ n(f,x)=E| f(S_ n/n)-f(x)|^ p,\quad \| D_ n(f)\|^ p_ p=\int_{I}D^ p_ n(f,x)dG(x), \] and \(V_ p(f)=\int_{I}\sigma^ p(x)| f'(x)|^ pdG(x)\), where G(x) is a distribution function on I. Under some conditions it is shown that \(n^{p/2}\| D_ n(f)\|^ p_ p\to C_ pV_ p(f)\) as \(n\to \infty\), where \(C_ p=2^{p/2}\Gamma ((p+1)/2)/\sqrt{\pi}.\) Asymptotic rate when f' and f'' exist is also established. Several examples are discussed.