IDENTIFICATION OF GENERATIVE STRUCTURES IN EMPIRICAL DATA†
George J. Klir · International Journal of General Systems · 1976
A hierarchy of epistemological levels of systems is introduced. It consists of source systems, data systems, generative systems, structure systems, and various levels of metasystems. The problem of structure identification is described as follows: Given a neutral data system, identify the best representation of this system by a directed structure system whose elements are associated with subsets of variables of the data system, A procedure for solving the problem is described. It consists of generating meaningful structure candidates for a given data system, analyzing each of them on the basis of the given data and within the framework of an overall sampling mask, and comparing the results with the data. A partial ordering is introduced for the structure candidates. For each particular number of variables in the data system, the structure candidates together with the partial ordering form a lattice. The ordering in the lattice makes it possible to reach the best structure system by analyzing only a fraction of all possible structure candidates.