Invariance principles for martingales and sums of independent random variables (Q1081196)

This paper contains improvements of the approximation theorems of Strassen's and Major's type in the martingale case [\textit{V. Strassen}, Proc. 5th Berkeley Symp. Math. Stat. Probab., Univ. Calif. 1965/1966, 2, 1, 315-343 (1967; Zbl 0201.499) and \textit{P. Major}, Stud. Sci. Math. Hung. 12, 161-167 (1977; Zbl 0431.60033)]. The main theorem is as follows: Let f be a non-increasing differentiable function such that for all \(x\geq x_ 0\), \(1/\log(x\vee e)\leq f(x)\leq 10^{-3}\), \(f(x)x(\log \log(x\vee e))^{-1/2}\uparrow\), \(g(x):=\log (x\vee e)/f(x)\uparrow \infty\) with \(xg'(x)\) bounded. Suppose that \(\{X_ n,{\mathfrak F}_ n,n\geq 1\}\) is a martingale difference sequence with finite second moments such that with probability one \(s^ 2_ n := \sum_{j\leq n}E(X^ 2_ j| {\mathfrak F}_{j-1})\to \infty\), \(| X_ n| \leq f(s_ n)s_ n(\log \log (s_ n\vee e))^{-1/2}\). Let, for \(t>0\) \(S(t) := \sum_{j\leq n}X_ j\) if \(s^ 2_ n\leq t<s^ 2_{n+1}.\) Then we can redefine \(\{X_ n,{\mathfrak F}_ n,n\geq 1\}\) on a possibly richer probability space on which there exists a standard Brownian motion \(\{B(t), t\geq 0\}\) and a random variable \(t_ 0\) such that with probability one \[ | S(t)-B(t)| \leq 10^ 3(f(t)t \log \log (t\vee e))^{1/2}\text{ for all } t\geq t_ 0. \]