Rate of convergence in the functional central limit theorem for semimartingales (Q1078911)

Let \((X_ n)_ n\) be a sequence of semimartingales on some standard filtration (\(\Omega\),\({\mathcal F},P,({\mathcal F}_ t)_{t\geq 0})\). Let X be a continuous semimartingale with independent increments (hence it has non- random characteristic). It is known [see e.g., \textit{J. Jacod}, \textit{A. Klopotowski} and \textit{J. Memin}, Ann. Inst. Henri Poincaré, Nouv. Sér., Sect. B 18, 1-45 (1982; Zbl 0493.60033)] that under some technical conditions, \(L_ T(P\circ X_ n^{-1},P\circ X^{- 1})\to^{n}0\), \(L_ T\) being the Levy-Prohorov distance on \({\mathcal D}([0,T]).\) The main purpose of this work is to give an estimate of that distance when the semimartingales \(X_ n\) have the most general form (theorem 1). We present the result (due to typographical difficulties) only for the case when \(X_ n\) are continuous and starting from 0. In this case \(X_{nt}=\alpha_{nt}+X^ c_{nt}\), \(X_ t=\alpha_ t+X^ c_ t\) where \(\alpha_ n\), \(\alpha\) are processes with bounded variation and \(X^ c_ n\), \(X^ c\) are continuous local martingales. Then, if \(\sup_{s\leq t}| \alpha_{ns}-\alpha_ s|\) and \(<X^ c_ n>_ t-<X^ c>_ t\) converge to 0 in probability, \[ L_ T(P\circ X_ n^{-1},P\circ X^{-1})= \] \[ O((E \sup_{t\leq T}| \alpha_{nt}-\alpha_ t|)^{1/2}+(E \sup_{t\leq T}| <X^ c_ n>_ t-<X^ c>_ t|)^{1/3})| \ln (E \sup_{t\leq T}| <X^ c_ n>\quad_ t-<X^ c>_ t|)|. \] The proof is interesting. It uses random change of time to embed a semimartingale into the Wiener process.