122 The Danger of Certain Correlation Formulae 
This discussion of the danger of Mr Yule's correlation formulae is by no means 
idle since the Coefficient of Association at least has recently been adopted by 
certain Italian writers. Thus Niceforo (" Contribute alio Studio della Variability 
di Alcuni Caratteri Anthropologic " : Atti della Societd Romana di Anthropologia, 
Vol. XVI. Fasc. 1) discusses the correlation between the mean and the coefficient 
of variability in twenty series of anthropological measurements and finds the 
Coefficient of Association to be — "909 while the actual value of the correlation 
coefficient, found by the product moment method from his data, is — "616. I have 
however been unable to reconstruct the fourfold table from which this coefficient, 
— '909, has been deduced so that there may be some slip in Niceforo's arithmetic. 
This is the more possible since out of the coefficients of variation given, very few 
as printed can be correct. Niceforo also states that the Coefficient of Association 
has been extensively used by himself in a paper on " Correlations e'conomiques " 
in the Journal de la Societe de Statistique de Paris, 1911, but I have not been 
able to find his paper in that journal. There is very little doubt that Mr Yule's 
" theoretical value of the correlation coefficient," owing to its apparently simple 
form, will also be seized upon by those who have not sufficient theoretical know- 
ledge to judge of its real value. In this way much harm will arise to statistical 
science, owing to the postulating of false conclusions resulting from the use of 
inadequate formulae. 
