A Note on the Classification of Human Physique by Principal Component Analysis

Abstract

(1) In order to make more comprehensive and statistically efficient indicators of a human physique, principal component analysis was applied to the anthropometric measurements. The subject used in this study comprised 200 Japanese female university students, who were measured on 30 items of measurements as well as photographed with three standardized pose. Further object of this report was to compare the results obtained with those derived from the data of 256 soldiers by MASUDA.(2) The interpretations of the principal components were made mainly on the basis of the coefficients of measurements in each component and also by reference to the photographs of the subjects. The first three principal components were interpreted as a size factor of a human physique, as a shape factor expressing lean-obese relation, and as a shape factor expressing shoulder shape. The percentages of the variances accounted for by each principal component to the total variance and the sum of them were approximately 41.0%, 16.3%, 6.6%, and 63.9%, respectively.(3) As to the meanings of the first two principal components, a perfect accordance was found between the present results and those of MASUDA. The interpretations were also confirmed by the result of HAMMOND, who had applied factor analysis to the data from 2967 children in Britain. The fact suggested the possibility of taking the first two principal components, that is, a size factor and a shape factor expressing lean-obese relation, as the best two indicators of a human physique irrespective race, sex and age.(4) As to the meaning of the third principal component, and as to the contribution ratios of each principal component to 30 measurements, the present results differed slightly from the results of MASUDA. The differences may be ascribed to the nature of the subjects and the initial choice of the measurements.