cc Tables for Statisticians and Biometricians [XXXVIII XLI 



An alternative is now open to us : the curve is a /-curve of Pearson Type I, but 

 if we use the first four moments we shall find the asymptote is not exactly at x = 0, 

 as we have assumed it to be when finding the abruptness-coefficients. We will 

 accordingly neglect /*& and fit our curve from the mean, /x 2 and /u, 3 . 



The equation to the curve referred to # = as origin is 



_ \ m * 

 b) ' 



and we find mi = - '062,235, m z = 1 24-570,905, 



and 6 = 192'594,54, or the equation to the curve is 



= 4052-866 



, 124-5709 



_ 



~ 



a92-5945/ V~ 192-5945/ 



We find # 2 for this curve =9-017,46, instead of the data value 8'946,97, a dif- 

 ference of little statistical importance, 



The graph of the curve with the data as a histogram is on p. cxcix, and it is 

 clear that the fit is good. 



If the frequencies be computed by the formula which expands an incomplete 

 B-function in terms of incomplete F-functions*, we have the following results: 



whence by Table XII of Part I, P = *94, indicating a very good graduation of the 

 data. 



* See Tracts for Computers, No. VIL p. 41. The formula is good when one power is very large as 

 compared with the other. 



