Factor data analysis and econophysics : application in market segmentation

Abstract

In this paper, an overview of factor data analysis methods is presented, as an alternative approach to classic statistical methods and it is shown that they are a powerful tool for analyzing economic phenomena. The principles on which data analysis methods are based are in a large degree inspired by physics, not only as general considerations but also as specific concepts, terminologies and methods. The notions of energy, entropy and inertia are matched with information theory, linear algebra and statistics to provide powerful tools for modeling and analyzing non-linear economic phenomena. Considering that any phenomenon under study is a complex open dynamic system where a large number of factors interact with each other, factor data analysis methods are able to examine such interactions as a whole, instead of a set of independent pair-wise comparisons of factors. The mechanism underlying these methods is to map the problem to a multidimensional vector space and based on the data themselves, to discover the underlying patterns, to find out how series of figures organize and which variables or group of variables are correlated. Model construction is thus not restricted to any initial assumption and is entirely driven by the data (Greenacre, 2007). In order to depict the potential of such methods in economic analysis, we present the application of multiple correspondence analysis to the market segmentation in the business plan for an internet radio venture.