APPRECIATION OF INTELLIGENCE 9 



association differ from the correlation coefficients calculated 

 on the same material may be seen from the following table, 

 which gives the coefficients of association ^ for various parts 

 of Dr. Warner's data, and the coefficients of correlation ^ 

 for the same data which had to be calculated by Mr. Yule 

 for his criticism of the L.C.C. investigation, 



I. TABLE SHOWING THE RELATIONSHIP BETWEEN MENTAL 

 DULLNESS AND BAD NUTRITION DEDUCED FROM {a) ASSO- 

 CIATION COEFFICIENTS AND (^) CORRELATION COEFFI- 

 CIENTS. 



This shows that the coefficient of association is about 

 50 % higher than the coefficient of correlation so far as this 

 material is concerned, but the value of the association co- 

 efficient depends on the choice of axes ; Mr. Yule in the 

 P/iil. Trans, paper already cited ^ gives a table for the special 

 case where the axes are taken at the mean values of the 

 variables showing a maximum difference of -i throughout 

 the whole range of the correlation coefficient. When the 



directly comparable with the correlation coefficient. He says (p. 300) : ' The 

 associations, however, are all high (very high compared with most coefficients 

 of organic correlation with which we have to deal), ranging from .794 to '952.' 

 No comparison, except possibly as to sign, can be made between the two types 

 of coefficients. 



* Extracted from Mr. Yule's Phil. Trans, paper, pp. 300, &c. 



^ Extracted from Mr. Yule's review in the Journal of the Royal Statistical 

 Society, vol. Ixiii, p. 549. 



^ Phil. Trans. Royal Society, 1900, p. 276. 



