How to (and How Not to) Analyze Deficient Height Samples

Abstract

Historians may make misleading inferences from historical height data based on military records because before universal conscription many armies had minimum height requirements. Are some methods of allowing for the missing small soldiers better than others? Two decades of experience in working with such deficient height distributions yield strategies to use and to avoid in order to obtain robust estimates. The author concludes that the quantile bend estimator procedure is to be avoided, because the other procedures available—the Komlos-Kim method, truncated ordinary least squares regression, and truncated maximum likelihood regression—are more robust and appropriate. He demonstrates the consequences of the choices among those methods by showing that they determine whether height evidence implies that the early Industrial Revolution in England led to an improving or a declining standard of living.

Keywords:

• anthropometric history
• Industrial Revolution
• living standards