scientific article; zbMATH DE number 3903651

From MaRDI portal
Publication:3681646





The upper estimates in approximation of the distributions of sums of independent random vectors with some special properties are given. Let \(F_ i\), \(i=1,2,...,n\), be probability distributions in k- dimensional Euclidean space \(R_ k\). Suppose that \(F_ i=(1-P_ i)U_ i+P_ iV_ i\), \(0\leq P_ i\leq 1\), where \(U_ i\) and \(V_ i\) are probability distributions and \(U_ i\) satisfies the condition: \(U_ i(\{x\in R_ k:\| x-y_ i\| \leq \tau \})=1\) for some \(y_ i\in R_ k\) and \(\tau >0\). A number of probability distributions \(D_ j\), \(j=1,...,6\) are constructed, which are close to the convolution \(F=F_ 1*...*F_ n\) in Lévy-Prochorov's metric \(\pi\) or in a metric connected with \(\pi\). In particular \(D_ j\), \(j=4,5,6\), are infinitely divisible distributions such that \[ \pi (F,D_ j)\leq \sum^{n}_{i=1}P^ 2_ i+C(\max_{1\leq i\leq n}P_ i+\tau (| \ln \tau | +1)),\quad j=4,5, \] \[ \pi (F,D_ 6)\leq \sum^{n}_{i=1}P_ i+C \max \{\tau (| \ln \tau | +1),\max_{1\leq i\leq n}P_ i\} \] where C is a positive number only dependent upon the dimension of \(R_ k\). These estimates are of the same form as those in the case \(R_ 1\) due to the author and \textit{T. V. Arak}, see Dokl. Akad. Nauk SSSR 267, 14-18 (1982; Zbl 0525.60017) and Teor. Veroyatn. Primen. 28, No.2, 333-353 (1983; Zbl 0517.60023).











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3681646)