Convergence rates in the law of large numbers for martingales (Q2277650): Difference between revisions

The author extends well-known results by \textit{L. E. Baum} and \textit{M. Katz} [Trans. Am. Math. Soc. 120, 108-123 (1965; Zbl 0142.148)] and others on the rate of convergence in the law of large numbers for sums of independent identically distributed random variables to general zero-mean martingales. A typical result of the paper is the following. Let \(S_ n=X_ 1+...+X_ n\), \(n=1,2,...\), be a zero-mean martingale, \(1/2<\alpha \leq 1\), \(p>1/\alpha\), \(\gamma\in (1/\alpha,2]\), \(q>(p\alpha -1)/(\gamma \alpha -1).\) It is shown that if \[ \sup_{n}\| n^{- 1}\sum^{n}_{j=1}E (| X_ j|^{\gamma} | S_ 1,...,S_ n)\|_ q<\infty, \] then \[ \sum^{\infty}_{n=1}n^{p\alpha -2} P(| S_ n| >\epsilon n^{\alpha})<\infty \] holds for all \(\epsilon >0\) if and only if \[ \sum^{\infty}_{n=1}\sum^{\infty}_{j=n}j^{p\alpha -2} P(| x_ n| <\epsilon j^{\alpha})<\infty \] for all \(\epsilon >0\).