24 THE MEASUREMENT OF VARIATION. 



simpler to separate all the numbers into two groups, one 

 containing all the values greater than the general mean, 

 and the other all those less than the mean. Then the 

 arithmetic mean error is half the difference between the 

 mean of each group. When, as in the present instance, 

 several of the values are identical with the mean, half 

 of them must be put in each group. The mean of one 



70 

 group is now = 8.75, and of the other group 



90 



g- = 11.25, and the arithmetic mean error is 



11.25-8.75 



2 



This method of estimating variability has frequently 

 been employed in recent times, especially in America. 

 Thus Minot * used it for comparing the variability of 

 guinea-pigs at various periods of their growth. Brew- 

 ster f used it for calculating the amount of varia- 

 tion in a number of head, face, and limb measurements 

 which were made by Weisbach J on individuals of 23 

 different races of men. In the general mean are in- 

 cluded the measurements of 195 individuals, represent- 

 ing 20 different races. The following are some of the 

 mean values for the arithmetical mean error, calculated 

 as percentages on the mean size : 



Nose length, 9.49 per cent. Head length, 2.44 per cent. 



" breadth, 7.57 * " breadth, 2.78 



" height, 15.2 ' Upper arm length, 6.50 



Forehead height, 10.4 ' Forearm length, 3.85 



Under jaw length, 4.81 ' Upper leg length, 5.00 



Mouth breadth, 5.18 * Lower " 5.04 



Foot length, 5.92 



* J. Physiol., xii. p. 138, 1891. 

 fProc. Amer. Acad. Arts Sci., xxxii. p. 268, 1897. 

 j Zeitschrift f. Ethnologic, Bd. ix. Supplement, 1878. 



