168 A BIOMETRIC STUDY OF BASAL METABOLISM IN MAN. 



ideal zero deviation, and BO far as this test is concerned must accord- 

 ingly be regarded as furnishing the best basis for predicting the metab- 

 olism of an unknown subject. 



Turn now to the average deviations without regard to sign. These 

 show the average error either above or below the actually observed 

 values. The averages are given in table 61. For the whole series of 

 64 individuals in which prediction is based on the averages per unit 

 in the Gephart and Du Bois selection 79 the average error is 100 calories 

 by the Du Bois height-weight chart as compared with 141 calories by 

 body-weight, or 6.08 per cent as compared with 8.57 per cent of the 

 average heat-production of the individuals tested. In predicting the 



TABLE 61. Average deviation without regard to sign of total heat-production as predicted 



from the mean heat-production per unit of body-weight or surface in 



standard series from the actual heat-production. 



metabolism of the 72 individuals of the Gephart and Du Bois selection 

 from averages based on the 64 other individuals, the average deviations 

 range from 87 to 106 calories, or 5.35 per cent for surface by the Meeh 

 formula, 5.46 per cent for surface by the Du Bois height-weight chart, 

 and 6.55 per cent for body-weight. Errors are much larger in the female 

 series, ranging from 6.98 per cent to 18.21 per cent, but with the order 

 of errors always lowest for prediction from body-surface by the Du Bois 

 height-weight chart, highest by body-weight, and intermediate in 



79 In working with the subgroups great irregularity must be expected because of the limited 

 numbers of individuals. In the case of the 17 individuals discarded from the original Nutrition 

 Laboratory series by Gephart and Du Bois the results of predicting from body-weight are partic- 

 ularly bad. The error is 6.43 per cent in the case of the height-weight chart and 13.79 per cent 

 in the case of body-weight. In the first supplementary series prediction from body-weight gives 

 slightly greater error than prediction from body-surface by the Meeh formula, but slightly less 

 error than prediction from the Du Bois height-weight chart. In all other series the error by the 

 height-weight chart is considerably less than by the body-weight method, and in all but two cases 

 it is less than prediction by the use of means for heat-production per unit of surface-area by the 

 Meeh formula. 



