On the rate of convergence in the multivariate CLT (Q806863): Difference between revisions

The author proves multidimensional Berry-Esseen theorems without Fourier techniques. First, a short inductive proof based on Stein's method [see \textit{C. Stein}, Proc. 6th Berkeley Sympos. Math. Statist. Probab., Univ. Calif. 1970, 2, 583--602 (1972; Zbl 0278.60026)] and leading to an explicit constant \(c(k)\) is given for a class of functions on \(\mathbb R^ k\) including indicators of convex sets. Then multivariate functions of independent random elements are treated, extending results of \textit{W. R. van Zwet} [Z. Wahrscheinlichkeitstheor. Verw. Geb. 66, 425--440 (1984; Zbl 0525.62023)] and \textit{K. O. Friedrich} [Ann. Stat. 17, 170--183 (1989; Zbl 0671.60016)]. The key is an auxiliary result which is of general interest in connection with asymptotic expansions in the CLT: It does not require truncation. Although for reasons of brevity it is proved here by standard methods, a similar version can again be obtained via Stein's method.