Multivariate data analysis:: quo vadis?: II.: Levels of data-modelling objectives and possibilities

In part I we presented an analysis of the situation facing multivariate data analysis at the turn of the millenium. As a direct analogy to object-oriented programming we presented similar ideas for multivariate data analysis, resulting in a proposal for a new paradigm for object-oriented data modelling (OODM), which is invariant w.r.t. data structures and practical data contexts. In this part II we give a first overview of tomorrow's meta-principles,ideas and stimulants for the implementation of multivariate OODM. This is exemplified by analysing the typical objectives behind data modelling, and we give numerous suggestions for what may be important development areas for the near future. We have arranged the many disparate data analysis objectives in a series of 'levels of modelling' in an attempt to make a systematic categorization. Copyright (C) 2003 John Wiley Sons, Ltd.