The Distribution of Products of Independent Random Variables
The Distribution of Products of Independent Random Variables
复制标题
DOI:
10.1137/0114046
复制
发表时间:
1966-03
影响因子:
1.9
通讯作者:
M. Springer;W. E. Thompson
中科院分区:
文献类型:
--
作者:
M. Springer;W. E. Thompson
MD SPRINGER AND W. E. THOMPSON 1. Summary. Fundamental methods are developed for the derivation of probability density functions (pdf’s) of products of n independent random variables (irv’s), and are used to obtain particular results which, aside from the case n 2, are believed to be new. The methods use the Mellin integral transform, and are a generalization to n variables of a method presented by Epstein [1]. Pdf’s are obtained inexplicit form for products of n monomial1, n-< 10 Cauchy, and n _< _ 7 Gaussian variables. Tables for products of n 2, 3, 6 Gaussian irv’s N (0, 1) have been calculated using this method [12], abridged versions of which are included in this paper. Entries in the unabridged tables were obtained with accuracy to six decimal places and permit linear interpolation with four-digit accuracy in the area column. These tables offer a heretofore unavailable tool to the engineer and research scientist concerned with reliability analysis, com-munications theory, and other applications requiringconsideration of products of irv’s.2. Introduction. Unlike the distribution of sums of irv’s, the distribution of products of more than two has received relativelylittle attention, and results which are available supply little useful information to the applied scientist. Epstein [1] has suggested a systematic approach to the study of products ofi. rv’s using the Mellin integral transform, but did not carry out the application for products of more than two. Levy [2] posed the question of a general theory of multiplication of irv’s and derived some results for products of two variables. In 1962, Zolotarev [3] began the con-struction ofa general theory of multiplication of irv’sanalogous to the theory of addition based on infinitely divisible distributions. His program was carried out in a sequence of theorems, stated without proof, which show both the similarity to, and difference from, the results for addition. Jam-bunathan [4] and Sakamoto [5], respectively, have derived the distribution of products of beta and rectangular irv’s. Other established results deal