^ Records of the Indian Museum. [Vol. XXIII , 



Coeff. of variation : — 



2nd group - 4"42 12 + *2i 13 



1st group = 4' 1 5 °4 ±' IQ - 8 3 



Difference = "27 08 +-28 97 



T/?e difference is in no case significant. 



Passing on to the other constants we get :— 



P-i '• — 



1st group = 2-14 66 5 1 ±' 20 4§ 



2nd group = 1*89 60 75 + - 18 09 





Difference 



- 0-25 05 76 + -27 33 



fV 



1st group 



- 0-28 75 92+/51 96 





2nd group 



- 1-29 90 10 + -43 14 





Difference 



= i-oi 14 18+/67 53 



u 4 : 



1st group 



12-26 96 01 + 3-04 55 





2nd group 

 Difference 



=s II-9I 27 23 + 2*32 19 





= o"35 68 78+ .38 30 



Pi'- 



1st group 



-_= -08 357 + -io 99 





2nd group 



=b -24 890 + -I2 04 





Difference 



- 16 533 + " 16 30 



/V 



1st group 



- 3"3 2 56 + '65 80 





2nd group 



= 2*66 27 + '26 09 



Difference "66 29 + 70 78 



We conclude that the first hundred measurements are not 

 significantly differentiated from the second hundred in any way. 

 Both represent "random" samples of the same general population. 



It should be noted however that the difference between the 

 two samples of hundred each, is of the same order as the probable 

 error of the difference. In one case viz. /';,, the difference is 

 actually greater than its probable error. This shows that 100 is 

 very nearly approaching the critical limit of <l fair (i.e. representa- 

 tive) sampling." [See section III, footnote 8, pp. 32-33]. 



There is grave danger 0/ samples of less than one hundred being 

 not representative in character (at least so far as the stature of 

 populations of the same order of variability as the Anglo-Indian is 

 concerned). The discussion on p, 40 Section IV. shows however 

 that two hundred is about the lower limit for safe inferences about 

 the general population. 



