A Practical Guide for the Creation of Random Number Sequences from Aggregated Correlation Data for Multi-Agent Simulations

This article describes a scalable way to initialise a simulation model with correlated random numbers. The focus is on the nontrivial issue of creating predefined multidimensional correlations amongst those numbers. A multi-agent model serves as a basis for practical demonstrations in this paper, while the method itself is interesting for an even wider audience within the modelling and simulation community beyond the field of agent-based modelling. In particular, we demonstrate how streams of correlated random numbers for different empirically-based model parameters can be generated when just given aggregated statistics in the form of a correlation matrix. An example initialisation procedure is demonstrated using the open source statistical computing software "R" as well as the open source multi-agent simulation software "Repast Simphony". We also provide a digression for NetLogo users.