INDIVIDUALS AND MEASUREMENTS CONSIDERED. 



59 



remember that in selecting our series for comparison we must choose 

 those of adult age in order to eliminate the influence of growth. Some 

 of the best studies on the correlation between stature and weight for 

 example, those of Boas c2 and of Boas and Wissler 63 on Toronto and 

 Worcester children, as well as the more recent investigation of Elder- 

 ton 64 on the stature and weight of Glasgow school children, carried 

 somewhat farther by Isserlis, 65 are therefore not available for our 

 present purposes. 



^ The correlations between stature and weight in our adults are given 

 in table 12. 



TABLE 12. Correlation between weight and stature and partial correlation between weight and 



stature for constant age in the several series. 



The partial correlations in which the influence of age is eliminated 

 have been computed from the formula 



T 



a' ws 



V T 



ws ' aw 1 as 



Vl-r Vl-i 



and placed beside the others for comparison. 



It is to be noted that correction for the influence of age has modified 

 the values of the constants very little indeed. They have sometimes 

 been slightly raised and sometimes slightly lowered by correction for 

 this factor. Age differences in the series can not, therefore, account 

 for any of the observed differences in correlation. 



62 Boas, Rept. U. S. Comm. Educ., 1896-97, p. 1541. 



63 Boas and Wissler, Rept. U. S. Comm. Educ., 1904, p. 26. 



64 Elderton, Biometrika, 1914, 10, p. 288. 



65 Isserlis, Biometrika, 1915, 11, p. 50. 



