Richard von Mises and the ``problem of two races'': a statistical satire in 1934 (Q997157)

The authors analyze a 1934 paper by von Mises regarding its mathematical content, its statistical philosophy, and its political background. Given a population, which is subdivided into two classes \(A\) and \(B\), and a numerical measure \(Z\) of a certain characteristic of this population, what are the statistical implications if there are \(x\) elements in \(A\), and \(m-x\) in \(B\) attaining the maximum value of \(Z\)? Von Mises derived methods for testing hypotheses on the distributions of \(Z\) on \(A\) and \(B\), respectively, as well as for analyzing Bayesian inference on the relations between these distributions. Von Mises' ill-suited illustration of his statistical methods by comparing the scientific talents of two races within the population of a European country (apparently Germany) on the basis of Nobel Prizes awarded to citizens of this country, gives reason to the authors for interpreting this application of the theory as a ``statistical parody''.