1922.] 



P. C. Mahalanobis : Analysis of Stature. 



79 



Hence 

 From Table II, 



D 0-4820 

 x = — = ' = 1-34 

 <j 0-3602 



J(i+o)= -go 98 773 



\J(l — a)= '09 01 227 



Thus nearly 9% of homogeneous samples will have a greater 

 Variability. The inclusion of the new Mediterranean series does 

 not affect our previous conclusion. 



The Variability of the Anglo-Indian sample though higher than 

 the Average is not excessively so and the difference is not statistically 

 significant. 



Indian Caste Variability. 



(a) Whole Series. 



I shall now consider the Coefficient of Variation of the Indian 

 Caste data of Risley. Omitting 3 tribes in which the sample 

 consists of only 2 individuals I get a total of 84 Castes and Tribes. 



Distribution of 84 Caste Coefficients of Variation. 



Group 



-2*1 



-2*2 



-2-3 



-2-4 



"2'5 



-2-6 



-27 



-2-8 



-2-9 



-3-0 



Frequency . . 



2 



I 



2 



c 



I 







1 



3 



8 



7 



Group 



-31 



-3-2 



-3 '3 



-3*4 



-3'5 



-3-6 



-37 



-3'8 



-3'9 



-4-0 



Frequency . . 



5 1 



12 



13 



7 



' 



6 



3 







5 



1 



Grouping by *i, I get 

 Mean Value of Caste Coefficient of Variation =3'2395 

 Standard Deviation of Coefficient of Variation = *3943 

 Anglo-Indian Coefficient of Variation =4-0672 



Anglo-Indian Difference = "8277 



■8277 



x — 



D 



— = - — —. = 2-099, 



a 



•3943 



From Biometric Table II, i(i + a) = -g8 21 356 



J(i — a)=*oi 78 644 



Only about two per cent of Indian Caste samples will show 

 greater variability. It seems therefore likely that the Anglo- 

 Indian sample is really differentiated from the Indian Castes in 

 showing a just significant degree of greater variability. 



It should be noted that the Caste Variability is much lower 

 than the non-Caste Variability. 

 We have 



Non-Caste Variability =3'57 00 ±' 368 

 Caste Variability 3*2395 + -0290 



Caste 



Difference = *3305±'0422 



