24 



EMPIRICAL STUDIES OF MEASUREMENT 



median ratio method gives results as close to the unlike-signs method 

 as does the Pearson method. The reader who will examine Table 

 XL in connection with the original 1 correlation tables in Biometrika 



il-43 *K-S8 SI- If 71-103 



Vol. 

 I., 

 I., 



L, 



I., 



II., 

 II., 

 II., 

 II., 

 III., 



FIG. 7. 



will find also that where the Pearson Coefficient r and the Median 

 Ratio r diverge at all widely it is the latter which better fulfils 

 Pearson's criterion of telling how much nearer the most probable 

 value of a second member of a pair is to the value of the first mem- 

 ber than it would be with no relationship at all. 



TABLE XI. 



Page Traits to be Related 



84 Longevity of adult brothers 

 126 No. of stamens with No. of pistils in 

 late flowers of Ficaria ranunculoides 

 214 Human head length with head width 

 216 Human height with left middle 



finger length 

 97 Capsule height of brother plants 



(Shirley poppies) 

 97 Stigmata of brother plants ( Shirley 



163 NoVof stamens with No. of pistils in 



lesser celandine from Surrey 

 498 Longevity of husband with longev- 

 ity of wife. Friends' records 

 170 Cephalic index of brothers 

 Average difference of r by Pearson Coefficient from r by cos wU .055. 

 Average difference of r by Median Ratio from r by cos irU .045. 



1 These examples are all taken from the first three volumes of Biometrika, 

 the ' Vol.' and ' Page ' of the table referring to that journal. 



