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



As may be seen, analyst A has a very symmetrical distribu- 

 tion of summations as 28*5 per cent of the analyses have sum- 

 mations above and 31*3 per cent below that of the maximum 

 group. His average summation of 100*11 per cent practical!}' 

 coincides with the general average, the constant error being 

 only 0*02 per cent. Also the proportion of analyses (4:0 - 2) falling 

 in the group of maximum frequency is high. In fact, having 

 regard for the large number of analyses made by this chemist, 

 this proportion may be considered as high as is obtainable in 

 ordinary practice. Analyst B has an evident constant minus 



Fig. 3. 



ff./Z ^f 37 



<?i?.62 ff.87 /00./2. ZOO.37 j '00.6 '2. S0O.J7 /0JJ2. 



error ; it amounts to 0*22 per cent, as his average summation 

 is 99 - 91. The constant error is, indeed, sufficiently large to 

 throw the group of maximum frequency to 99"75-99 - 99 per 

 cent. The distribution curve of B is somewhat asymmetric 

 and a larger proportion of summations lie above the group of 

 maximum frequency than below; the percentages are, respec- 

 tively, 37*8 and 25*0. The proportion of summations (364) in 

 the maximum group is good. Analyst C has a constant plus 

 error of 0*12 per cent, his average summation being 100*25. 

 His group of maximum frequency is 100 , 25-100*49 per cent, 

 but the group 100*00-100*24 per cent contains nearly as large 

 a proportion of analyses. The actual maximum for the fitted 



