A CRITIQUE OF THE BODY-SURFACE LAW. 



173 



in three ways : by the determination of the mean error with regard to 

 sign, by the determination of the mean error without regard to sign, 

 and by the determination of the square root of mean square deviation 

 of the predicted from the actually measured values. 



The mean deviations with regard to sign appear in table 66. With 

 one exception they indicate that in the nine comparisons with the 

 three subseries (I-III) prediction from the constants of the Gephart 

 and Du Bois selection is on the average too high. This is also true of 

 the whole series of 64 individuals. The actual amount of the deviation 

 is not large. It ranges from 3.6 to 38.2 calories in the subseries and 

 from 2.5 to 9.6 calories in the combination series. In terms of per- 

 centages of the mean heat-production of the groups dealt with these 



TABLE 65. Differences in calories between square root of the mean-square errors of prediction 

 by use of means and by use of straight-line equations. 



average deviations with regard to sign range from 0.15 to 2.25 per 

 cent, but only 2 of the subseries show a percentage deviation of over 

 1 per cent, and the 3 constants for the whole series of 64 individuals 

 show a deviation of less than 0.6 per cent. 



Since the constants based on the Gephart and Du Bois selection 

 give slightly too high results when used to predict the heat-production 

 of other individuals, it is necessary that the constants of this other 

 series give values which are too low when they are used to predict the 

 heat-production of the individuals of the Gephart and Du Bois selection. 

 We note, therefore, that the average deviations for the predicted values 

 of the Gephart and Du Bois selection are negative in sign throughout. 

 The actual values are roughly comparable with those already con- 

 sidered, ranging from 4.1 to 7.4 calories, or from 0.25 to 0.46 per cent 

 of the mean heat-production. 



This difference in the sign of the average deviation in the two 



