H. H. Robinson — Chemical Analyses of Igneous Rocks. 263 



This unit has been found by taking different sized units, 

 from necessity, as multiples of 0*05 per cent, and determining 

 which gives the best distribution. It is thus found that a 

 unit of 0*15 per cent most nearly meets the requirements for 

 the entire range of summations. The units of 0*20 and 25 

 per cent also produce a symmetrical distribution between the 

 restricted limits of 99*50 and 100'74 per cent, which is better 

 than that given by the 0*15 per cent unit. As three-quarters 



Fig. 1. 









450 

 ( 



> 















420- 

















J39t> 

















/ 360- 

















f 330- 

















7 300- 



o\ 















270- 



o\ 















240- 

















210 





\° 















180 

 750 



/ZO 

 90 

 60 

 30 





\° 



\°0 





. oooipoairwoocpdt^ 



i I i — i — \ 



— | — i — i- 



> i 



— i — i — r 



i i 



1 ! 1 1- 



, ^0?rf 5 9o™co<pOrfitoo < r tnlf 



<)7d7 9S32 



177 9922 9967 IO0J2 10057 IO/.02 IOIA7 I0L<)1 10237 102.82 



of all summations fall within the foregoing limits there is 

 practically little choice between the 0*15, 0*20, and 0*25 per 

 cent units. A study of a larger number of analyses than here 

 used might show the 0'15 per cent unit to be less suitable than 

 the 0-20 or possibly the 0*25 per cent unit. The 0*15 per 

 cent unit has been used to show the distribution of summa- 

 tions in general and the 0'25 per cent unit has been chosen, 

 purely as a "matter of convenience, to show the distributions 

 for individual analysts which are later given. 



For the unit of 015 per cent there are three possible com- 

 binations of the 0*05 per cent unit. The values for these three 

 combinations have been plotted, as is customary, and a sym- 

 metrical curve has been fitted to them, as shown in fig. 1. In 

 this figure the ordinates are the actual number of analyses, but 



