Closeness of convolutions of probability measures (Q605035): Difference between revisions

In considering approximations of distribution functions of sums of independent random variables, the accuracy of the approximation may be improved by using the convolution of identical factors when the terms in the sum are not too different. In this very technical paper, the author derives an explicit bound for the total variation distance between two convolution products of \(n\) probability distributions, in one of which all the factors are identical. Approximations by finite signed measures are also considered. The main results are given in Theorems 2.1 and 2.2; where a Banach algebra technique for measures on a measurable abelian group are used. The bounds improve those by the same author [Theory Probab. Appl. 46, No. 1, 103--117 (2001; Zbl 0994.62006)] and by \textit{W.-L. Looh} [Ann. Appl. Probab. 2, No. 3, 536--554 (1992; Zbl 0759.62007)]; they depend on a ``magic factor'' [see \textit{A. D. Barbour, L. Holst} and \textit{S. Janson}, Poisson approximation. Oxford Studies in Probability. 2. Oxford: Clarendon Press (1992; Zbl 0746.60002)].