188 METABOLISM AND GROWTH FROM BIRTH TO PUBERTY. 



there seemed to be a somewhat wider dispersion of points about the 

 curves for surface than about the curves for weight. For an exact 

 analysis of, (1) the suitability of these curves for prediction, and (2) of 

 the relative merits of surface and weight curves for prediction, mathe- 

 matical treatment of each child on the two bases of weight and surface 

 referred to total calories is essential; consequently we have prepared 

 two tables for the boys and girls, respectively, giving the essential 

 physical data for the individual children, and the values for both the 

 measured heat and the heat predicted on the two bases of weight and 

 surface. Since we are testing the accuracy of our general curves for 

 the prediction of heat, the differences between the predicted and actual 

 heat are given in these tables for both methods of prediction on the 

 basis of the predicted less the actual, so that if the prediction is lower 

 than the actual the difference will be indicated by a minus sign. The 

 percentage differences are also given, since percentage values can 

 alone be used for the compariison of children of various sizes. 



In addition to the data for the predicted heat on the bases of weight 

 and surface, we have employed a third method of prediction for the 

 boys, in which use was made of the multiple prediction formula of 

 Harris and Benedict 1 as proposed by them for the prediction of the 

 most probable total heat production of men. Exactly the same 

 treatment with regard to the differences between predicted and 

 actual, as well as the percentage differences, is accorded this method 

 of prediction. Thus we have in the tables the actual total heat 

 production per 24 hours of each child and then (by the three methods 

 of prediction) the most probable heat production of a child of similar 

 weight or (in the case of surface) of similar surface, taken from the 

 corresponding curve. The differences of the actual from the smoothed 

 curve values are then noted and recorded and the percentage values 

 found. Since with the first two methods of prediction, i. e., those 

 from curves, the smoothed curve is nothing more than an attempt to 

 equalize all the inequalities in the values, theoretically there should 

 be practically the same number of plus and minus differences in the 

 whole series, and in any event the sum of the plus or minus differences 

 should practically equal zero. 



PREDICTED HEAT FROM TOTAL CALORIES REFERRED TO WEIGHT 



(BOYS). 



The predicted heat for boys, based upon the curve for total calories 

 referred to weight (see fig. 26) is given in table 32 and may first be 

 considered. A superficial inspection of the table shows an approxi- 

 mately equal distribution of plus and minus signs. While this would 



1 Harris and Benedict, Carnegie Inst. Wash. Pub. No. 279, 1919. In this monograph the differ- 

 ences were computed by deducting the predicted values from the actual, instead of the actual 

 values from the predicted as in the present report. 



