In this paper the author proposes a method which will allow simple and fast
generation of two sets of random variables with given probability mass fun
ctions (PMFs) and crosscorrelation sequences (CCS). It is based on the comp
osition of three sets of independent and identically distributed discrete r
andom variables, two of which are correlated with each other only when the
lag is null. The method is applicable in cases where, once experimental mea
surements have established the statistical characteristics of two stationar
y random processes in terms of PMFs and CCS, it is necessary to implement a
simulation model, that is, it is necessary to generate random discrete var
iables whose PMFs and CCS match the empirical histograms. A case study is g
iven to apply the method proposed to modelling the relationships between th
e monomedia streams making up a multimedia stream. (C) 1999 Published by El
sevier Science B.V. All rights reserved.