PREDICTION OF THE BASAL METABOLISM OF YOUTH. 195 



our attention first to the actual differences as shown in column j, we 

 find that they range from to as high as 191 calories; but realizing 

 the very great differences in the actual heat production, we again 

 note that percentage differences alone can be considered for com- 

 parative purposes. Exactly the same reasoning with regard to the 

 effect of the curve in smoothing out plus and minus differences in 

 the individual points obtains here, namely, that we would expect to 

 find essentially the same number of plus as minus values, and a super- 

 ficial inspection of the table bears this out. 



On the basis of weight, 7 instances were noted in which the per- 

 centage difference was 20 per cent or more, while the heat based on the 

 surface gives but 5 instances. All but one of these are with children 

 whose body-weight is 10.7 kg. or below. This observation has two 

 important features: First, it confirms the view previously expressed 

 that the error of prediction is largest with the children of small weight. 

 Second, it implies that, so far as extreme errors are concerned, the 

 prediction by surface is somewhat better than the prediction by weight. 

 It is, however, only after a consideration of all the data that one can 

 draw final conclusions. With all the series of boys the average per- 

 centage deviation of predicted from the actual is 7.5 per cent. 

 (See table 33.) Thus it appears that while the number of cases of 

 extreme error was less when the measured surface was used in pre- 

 dicting the total calories, on the average this method of prediction 

 gives slightly greater error than when the calories are referred to 

 weight. 



Our general impression that the error of prediction is greater with 

 children with smaller surface is sufficiently substantiated to justify 

 our making a division of the values and considering boys with a surface 

 area below 0.45 square meter in one group and those above this area 

 in another. This division will be not unlike the comparison on the 

 basis of weight, since a body-weight of 10 kg. corresponds approxi- 

 mately to a surface area of 0.45 square meter. The results for this 

 division are also given in table 33 and show that with the boys below 

 0.45 square meter of surface, the error of prediction is 7.7 per cent, 

 while the boys above this area have a slightly better percentage, 

 namely, 7.3 per cent. It thus appears that with the smaller boys 

 the error of prediction is somewhat greater by weight than by surface, 

 i. e., 8.7 per cent as compared with 7.7 per cent. This fact, taken 

 alone, would imply a greater correlation between body-surface area 

 and heat production than between body-weight and heat production. 

 On the other hand, with the boys weighing 10 kg. and above and with a 

 surface area of 0.45 square meter and above, we find that the error 

 of prediction is 1 per cent greater when surface area is used that is, 

 7.3 per cent for prediction by surface as compared with 6.3 per 

 cent by weight. With boys with a body-weight of 10 kg. or above, 



