IRE Simulation, and simulation experiments chapter Michael D. Heine Butterworth & Company Karen Sparck Jones All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, including photocopying and recording, without the written permission of the copyright holder, application for which should be addressed to the Publishers. Such written permission must also be obtained before any part of this publication is stored in a retrieval system of any nature. I0 Simulation, and simulation experiments Michael D. Heine 10.1 Introduction The term `simulation' is variously defined: `Some simulations seem to be searches for structure, for important variables, for hypotheses. Others are tests of hypotheses. Still others trace the implications of a theory or simply enumerate the consequences of particular mechanisms, policies, procedures, or models.' (Chapman p.482) As the terms `system' and `model'[OCRerr]ven `information'-which tend to occur with `simulation' are also variously defined, the difficulties in communication are compounded. Since we have yet to agree to a system of thought that encompasses all human cognition, a degree of ambiguity is perhaps inevitable, given the very abstract connotation that simulation can have. A brief classification of meanings is as follows. To some authors the term `simulation' is associated with an essentially verbal analysis of an area of phenomena of interest-which is termed `the system.'. Or the `system' may be a sub-area of the area of interest, delimitation of the sub-area being seen as an intrinsic part of the analysis. (The verbal description may be accompanied by diagrams representing, in an indicative rather than a rigorous way, the thoughts in the analysis.) To other authors, simulation is a branch of applied mathematics (or is all of applied mathematics), the concern being to represent real entities and relationships (i.e. observables) by numeric variables and analytical functions. Here simulation is an embodiment of the conventional approach of the physicist or engineer, and the definition is little different from that of classical science itself. `Objectivity' in analysis is recognized-indeed is required to be so. The critical thing to yet other authors is that the techniques used to explore mathematical relationships are implemented by computers, often involving exploration of the cumulative effects of manipulating values taken randomly from a (0,1)-uniform continuous distribution. Here no particular concern is shown for either the problem of system delimitation or the aims of the study. The apparent vagueness in `the simulation approach' just described is a little illusory, for beneath the various definitions can be discerned general 179