PREDICTION OF THE BASAL METABOLISM OF YOUTH. 189 



tend to indicate that the differences between individuals were in the 

 main eliminated in drawing the curve, special consideration should be 

 given to the numerical and percentage differences of the predicted less 

 the actual heat. The numerical differences are expressed in column g 

 of table 32 and vary from to as high as 177 calories. Since there 

 was a very wide variation in the heat actually measured, i. e., from a 

 minimum of 163 calories with two of the youngest boys to as high 

 as 1,401 calories with the oldest boy, the percentage differences alone 

 are of value for comparative purposes. The average percentage differ- 

 ence for the whole series of boys, as given in table 33, indicates that 

 the total heat can be predicted from the curve for calories referred 

 to weight with a deviation of 7.4 per cent. Although an average 

 error of this magnitude may be considered as large, it must be remem- 

 bered that the data for growing children are extraordinarily few. 

 Hence, if we are to take advantage of this series of observations, the 

 hint given by the general trend of the line in figure 26 may on the whole 

 be regarded, provided that too much emphasis is not placed upon this 

 method of prediction. 



On inspection of the percentage deviations in table 32, we find that 

 the differences of the predicted from the actual are rather considerable 

 in some instances, there being deviations with 7 boys of 20 per cent 

 or more. In all of these seven instances the weight of the boys is 

 10.7 kg. or below, suggesting that the error of prediction with small 

 children is much greater than it is with the larger children. Appar- 

 ently some weight not far from 10 kg. represents an approximate 

 dividing-line between a reasonably close prediction and a much more 

 gross prediction. If we use 10 kg. as an arbitrary dividing-line, and 

 calculate the average deviation of the predicted heat from the measured 

 heat for children weighing 10 kg. and above, we find this average 

 deviation to be 6.3 per cent. (See table 33.) For the boys under 

 10 kg., we find the average percentage deviation to be 8.7 per cent. 

 Thus, what is obvious to the eye, both with regard to the dispersion 

 of points and to the general magnitude of the percentage differences in 

 column h of table 32, is amply verified by a calculation of the average 

 percentage deviation of the predicted values from the actually observed 

 values. 



That some other weight than 10 kg. may not be slightly better, 

 mathematically, for a division-line is not disproved, but for practical 

 purposes this division seems t to be all that the calculations justify. 

 To test the theory that some other weight might be better for division 

 purposes, the percentage differences have been computed for boys 

 w-eighing 15 kg. and above. On this basis we find that the prediction 

 error is slightly less that is, it now becomes 5.8 instead of 6.3 per 

 cent. Since, however, we purpose predicting the heat, not from a 

 formula, but by drawing points from this curve, the lessening of the 



