172 A BIOMETRIC STUDY OF BASAL METABOLISM IN MAN. 



In comparing the results of predicting total heat-production from 

 body-surface by equations and by considering it proportional to the 

 average heat-production per square meter of body-surface, we note 

 that the differences are far smaller than those found when body-weight 

 is used. It is not, therefore, so essential to use the equations when 

 body-surface is to be employed as a basis of prediction as when body- 

 weight is used. But in predicting from body-surface the equations 

 give better results in 8 out of the 14 comparisons. 



Table 65 gives the comparison of the square root of mean square 

 deviation of the calculated from the actual values for the prediction 

 by the use of means only and by the use of linear regression equations. 

 In prediction from body-weight, the straight line gives far more satis- 



TABLE 64. Differences in calories between the average deviations without regard to sign 

 resulting from the use of means and straight-line equations for prediction. 



factory results. In the case of the two body -surf ace measurements 

 there is less difference. It is important to note that in the case of the 

 Du Bois height-weight chart, in which body-surface is not merely a 

 function of weight, the evidence for accuracy of prediction is in favor 

 of the linear prediction formula. This is shown by the fact that in 

 6 of the 7 cases prediction from the mean heat-production in the 

 standard series gives a larger square root of mean square deviation 

 than prediction by the use of linear equations. 



Taking all the three lines of evidence together, a material superiority 

 of the linear regression equation over the method heretofore used for 

 purposes of prediction is evident. 



We now turn to a comparison of the results of predicting metabo- 

 lism by means of straight-line equations based on body- weight and 

 based on body-surface. We shall compare the results of such prediction 



