IQ22.] 



P. C. Mahalanobis : Analysis of Stature. 



81 



To sum up, 



77&£ Anglo-Indian Variability is significantly greater than Caste 

 Variability but is not beyond the range of homogeneous Variability. 



Other Comparisons. 



I shall give a short summary of other comparisons, reserv- 

 ing a fuller discussion for a future paper on the Caste data 



Pooling together the 84 Caste and the 109 other samples we 

 get a total of 193 (all samples). 



I find 



Mean Value of Coefficient of Variation = 3*4231 + '0240 

 Anglo-Indian Co-efficient of Variation =4*0672 



Anglo-Indian Difference = '6441 



Standard Deviation - -4949 + -0169 

 6441 



Thus 

 From Biometric Table 



x — 



1-301, 



'4949 

 £(i±a) = '90 31 995 



J(i-a)=09 68 005. 



Anglo-Indian Variability would be exceeded by nearly 10% of 

 total (Caste and non-Caste) samples. 



Selecting samples greater than 25 we get a total of 137 fairly 

 reliable samples. 



Distribution of ij? Selected Coefficients of Variation. 



Group 



Be- 

 yond 



2*2 



-2-3 



-2*4 



-2'5 



-2'6 



-27 



-2'8 



-2 -9 



-3-0 



-3 1 



Frequency . . 



I 



I 







I 







3 



4 



8 



10 



5 



Group 



-3'2 



-3'3 



-3*4 



-3 :> 



-3-6 



-37 



-3'8 



-3*9 



-4-0 



-4-1 



Frequency . . 



13 



H 



11 



21 



12 



9 



8 



9 



1 



4 



Group 



-4*2 



-4*3 



Total. 

















Frequency . . 



I 



1 



137 





Grouping by "i I find : — 



/x 2 = i4'4 2 I2 2 9 



f*8= - *7'9 8 7 2 9° 

 M4 = 760*82 26 19 

 Hence 



£,= -10796 ±'14439 



fo= 3'6589 2 ±'9 I 379 



Thus we are justified in applying the normal integral for cal- 

 culating the chances for any deviation. 



