Berry-Esseen bounds for standardized subordinators via moduli of smoothness (Q2641423): Difference between revisions

Let \((X(t))_{t\geq0}\) be a subordinator, that is, a real-valued random process starting at the origin, having independent stationary increments and right-continuous paths. Further, assume \(E(X(t))=t\) for \(t>0\). Then \[ E(e^{i\zeta X(t)})=\exp\biggl(tE\,{e^{i\zeta T}-1 \over T}\biggr),\quad\zeta\in\mathbb R,\;t\geq0, \] for some nonnegative random variable \(T\). Assume \(0<\sigma^2=E(T)<\infty\) and set \(Z(t)=(X(t)-t)/\sigma\sqrt{t}\) for \(t\geq1\). The authors establish bounds for the uniform distance between the distribution function \(F_t\) of \(Z(t)\) and the standard normal distribution function \(\Phi\). These bounds involve moduli of smoothness of order 1 and 2. The results are extended to standardized Lévy processes with finite variance.