4 2 Records of the Indian Museum. [ Voi,. XXIII 



The curve is not significantly skew. But there is distinct 

 tendency towards lepto-kurtosis. 



The curve belongs to Type IV of Pearson's Skew Curves 1 

 The probable errors of /J, and a are quite large, and we mav 

 investigate whether the /?, -& " probability ellipse " touches the 

 Gaussian point G? 



In order to do this we must find 2, and 2 2 , the semi-minor 

 and semi-major axis of the " probability ellipse." 



A = 'o6 8 7 56 



:. = 3"5 i-i77 v iV.S, = 1-4 + -37 912 x [-4] = 1-551648 



= 3-6 X5 + -37 QI2 X[ . 51 = I " 688 5 



_ ' I 379 I 2 



A = 3'5o 46, 1-177 VN2 X = 1-55 79 91 



Similarly 1*177 V^S^ 13*51 71 2 



Multiplying by *i = -04769, we get 



semi-minoi axis = '0743 

 semi-major axis = '6446 



Tracing a probability ellipse with these values and centering 

 the ellipse at the point /?, = -07 and & = 3-5 approximately on 

 the diagram on p. 66 of Biometric Tables, we find that the Gaussian 

 point G falls just within the ellipse. We also note that the ellipse 

 covers a small area of the Type III region. 



We conclude therefore that a Gaussian distribution itself 

 is not unlikely and may be expected to give a good fit. Type 

 III is not altogether impossible but as the major portion of the 

 ellipse lies within the Type IV region, the lepto-kurtosis is prob- 

 ably just significant. 3 



Comparative Data. 



f Our frequency curve is approximately Gaussian in type* 

 The asymmetry is very slight, skewness is small and positive 

 (Mode is greater than the Mean) and the curve belongs to Type IV 

 with lepto-kurtosis. 



A. O. Powys 4 lias discussed distribution of stature for 

 different age groups of New South Wales criminals. The author 

 saj 'by Looking at the curves, we see that the material is 

 extremely homogeneous 6 .... the stature distribution of these 



■ See Memoirs cited above in footnote on p. 10. 



I A discussion of these points is given by A. Rhind, "Additional Tables 

 and Diagram or the Determination of the Errors ol rypeol Frequency Distribu. 

 Btometrtka \ ol. 7 (1910), p. j86 — 397. 



, y _ lU ' •'>""•"■">• is very slight and the distance between the Mode and 



house between 

 improved fit. 



1 (19 p. jo. """" l ' °> omet r ika V " L 



fbid., p. j8. 



• Ihe asymmetry is very slighl and the distance between th< 



he Mean is afso ouite small. On the whole there is very little to eh 



normal and a Ty pe l\ curve. The latter may give slighth in 



°' ' ow y s: "Anthropometric Data iron, Australia," Biot 



