Introduction

Abstract

It is worthwhile noting that — although sometimes known under different names — certain features of systems¹⁾ and of system modelling appear to be the same in different fields of research. The stability of a system is such a feature arising in almost every field of research, e.g. in biology, physics, economics, cybernetics, etc. The description of a system in terms of variables acting upon the system, called stimuli in psychology, predetermined variables in econometrics and input variables in system engineering and variables describing the response of the system, called endogenous variables in econometrics and output variables in system engineering is an example of a system modelling feature. It is not surprising, that in a variety of cases, statistical models, although from different fields of research, have the same structure. In order to give an impression of such a structure equivalence, a ‘basic’ decomposition y;(k) = s(k) + v(k) is considered. Here y(k) denotes some observable random vector which is decomposed into a systematic part s(k) and a non-systematic part v(k), for instance v(k) is a normally distributed random vector, uncorrelated with v(s) for k ≠ s, to be referred to as white noise. The index (k) may be time, regions, persons, etc. In engineering y(k) may be a signal s(k) corrupted with white noise. In econometrics, the decomposition is called a linear regression equation, when the systematic part s(k) is modelled as s(k) = x(k)ß, in which x(k) is some given row vector of exogenous variables and ß is the parameter vector.