A CRITIQUE OF THE BODY-SURFACE LAW. 



191 



the average deviation of the calculated total heat-production with 

 regard to signs, the signs of the constants in the first column of table 70 

 and in the first column of table 73 are disregarded, and the differences 

 represent merely the difference in the numerical magnitudes of the 

 discrepancy between observation and prediction. 



Considering the values in table 73, we see that in some cases the 

 equations involving weight, stature, and age give closer and in some 

 cases slightly wider average deviations above or below the true value. 

 In the larger series (IV-VII and total men and women) the equations 



TABLE 73. Comparison of average deviation (in calorics, with regard to sign) from actual, caloric-output 

 of heat-production calculated on the one hand from multiple regression equations invoking stature, 

 body-weight, and age and on the other from (a) the mean heat-production per unit of body-weight and 

 body-surface by Du Bois height-weight chart, from (6) the regression of total heat on body-weight and 

 on body-surface by the Du Bois height-weight chart, and from (c) the regression of total heat-production 

 on stature and body-weight. 



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

 ceding tables as follows: column II from III of table 60; column III from III of table 66; column IV from I 

 of table 60; column V from I of table 66; column VI from I of table 70. 



which take into account weight, stature, and age give somewhat better 

 results than those in which prediction is made by the other methods 

 employed. 



The figures set forth in tables 74 and 75 are so striking that they 

 require but few words of discussion. Consider table 74 showing the 

 average deviations without regard to sign of the calculated from the 

 actually determined heat-productions in the several series of individuals 

 when the former are computed in various ways. With one single and 

 numerically insignificant (+0.7 =0.04 per cent) exception the 45 differ- 

 ences are negative in sign, showing that the error of prediction is smaller 

 when multiple regression equations involving weight, stature, and age 

 are used than when any of the other 5 methods of estimating the heat- 

 production of a subject is employed. In the larger series (IV-VII and 



