122 



THE AMERICAN NATURALIST [Vol. XLII 



cerned with the problem of determining correlation from ma- 

 terial arranged in ranks. The most obviously simple way of 

 dealing with any mass of statistical material is to arrange the 

 individuals in a long row in order of ascending magnitude of 

 some character. It is not even necessary that the character be 

 accurately measurable on a quantitative scale to make such a 

 row. The second, and on the whole most important of Pearson's 

 new methods, shows how to deduce true variate correlation from 

 the correlation of ranks. The method is less accurate than the 

 usual product moment method. The third method deals with the 

 determination of variate correlation from the positive difference 

 of ranks. Regarding these last two methods Pearson's general 

 conclusion is "that variate correlations found by ranks may 

 prove to be a useful auril mri/ method of dealing with correlation, 

 when it is needful to give a rough answer to a problem in a brief 

 time, or when the material itself is incapable of bring accurately 

 measured. ' ' 



In connection with this matter of the analysis of statistics by 

 the method of ranks or grades it is of interest to note that Galton 

 has recently published a table 7 of deviates of the normal curve 

 which will be found useful in practical statistical work. In the 

 text accompanying the table (which is calculated by W. F. 

 Sheppard) the use of the method of grades in dealing with ma- 

 terial not susceptible of exact numerical measurement is dis- 

 cussed. 



Of value for practical biometrical computing work is an edi- 

 curve fitting. ' P 



