Models of statistical distribution

Abstract

This article focuses on a historical shift in the scientific attention given to two statistical models, the ‘normal’ distribution, which dominated ‘parametric’ statistical inference for a century from the 1880s, and the ‘power-law’ distribution, which is now seen as being much more commonplace than it once was. Each measures the relationship between the size and frequency of a variable, but the assumptions brought to that process differ considerably. Where one emphasizes equality and a typical form, the other highlights inequality and no form is representative. Where one is bounded, static and random, the other is open, dynamic and may reveal preferences. The analysis of social networks provides a case study. Drawing on Durkheim and Mauss’s essay on ‘primitive classification’, it is argued that a shift from national to world society may partly account for the relative prominence attached to these two models over time.