A CRITIQUE OF THE BODY-SURFACE LAW. 



187 



measured heat-productions for the group under consideration, the 

 differences in the errors of prediction range from 0.00 per cent to 

 0.29 per cent. 



If the test be based upon the whole series of men and of women 

 we find that the multiple regression equations give better results in 

 every case but one. In this case prediction from the linear equation 

 for total heat on body-surface area gives a mean deviation 0.2 calorie 

 per day less in the men than the multiple regression equations. This 

 represents a difference of 0.01 per cent only. 



The comparison on the basis of square root of mean-square devia- 

 tion is made in table 72. The results show that in 6 of the 8 larger 

 series (IV-VII) in which prediction is made from constants based upon 



TABLE 71. Comparison of average deviation (in calories, without regard to sign) from the actual caloric-output, 

 of heat production calculated on the one hand from multiple regression equations involving body-weight and 

 stature and on the other from (a) the mean heat-production per unit of body weight and of surface by the 

 Du Bois height-weight chart and from (b) the regression of total heat on body-weight and on surface area 

 by the Du Bois height-weight chart. 



* The differences in these columns are obtained from the first column of this table and the entries of preceding 

 tables as follows: column II from III of table 61; column III from III of table 67; column IV from I of table 61; 

 column V from I of table 67. 



a different group the error of prediction is greater by the equations 

 here being tested than by prediction from body-surface by the Du Bois 

 height-weight chart. The difference between the two methods is, how- 

 ever, very slight. In working units, it ranges from 1.1 to 4.7 calories 

 per day. In terms of percentages of the average daily heat-production 

 of the series of individuals dealt with, the differences in the errors of 

 estimation by the multiple-regression equations and the prediction 

 method based on body-surface range from 0.04 to 0.33 per cent. 



Turning to a comparison of the various methods of calculation 

 when the whole series of men and women are used, it appears in every 

 case except one that the multiple regression equations give the more 

 accurate prediction of metabolism. 



