18 CLARKE— STATISTICAL METHOD [April i8 



igneous rocks from British localities. Many of these analyses were 

 incomplete, especially with respect to phosphorus and titanium. 



F. An average of i,8ii analyses, from Washington's tables.^ 

 Calculated by H. S. Washington. The data represent material from 

 all parts of the world. 



Now, omitting minor constituents, which rarely appear except 

 in the more modern analyses, these averages may be tabulated to- 

 gether, although they are not absolutely comparable. The compari- 

 son assumes the following form : 



A. B. c. D. E. F. 



Clarke. Clarke. Clarke. Clarke. Harker. Washington. 



SiOo 58.59 59.77 59-71 60.91 58.75 58-239 



ALO3 15.04 15.38 15.41 15.28 15.64 15-796 



Fe.Oa 3.94 2.6s 2.63 2.63 5-34 3-334 



FeO 3.48 3.44 3.52 3.46 2.40 3.874 



MgO 4-49 4-40 4-36 4-13 4-09 3-843 



CaO 5.29 4.81 4.90 4.88 4.98 5.221 



Na.,0 3-20 3.61 3-55 3-45 3-25 3-9^2 



K.jO 2.90 2.83 2.80 2.98 2.74 3. 161 



H2O at 100° ^ .... .... .41 I -3^3 



I 1.96 [ 2.23 



HoO above 100° J 1.51 1.52 1.49 ) 1.428 



TiOa 55 -53 -60 .73 .12 1.039 



PoOs .22 .21 .22 .26 .02 .373 



99.66 99.14 99.22 100.61 99-56 100.583 



Although these six columns are not very divergent, there are 

 differences between them which may be more apparent than real. 

 Dift'erences of summation are partly due to the omission of minor 

 constituents ; but the largest variations are attributable to the water. 

 In two columns hygroscopic water is omitted ; in two it is not dis- 

 tinguished from combined water; in two a discrimination is made. 

 By rejecting the figures for water and recalculating to 100 per cent., 

 the averages become more nearly alike, as follows : 



^ U. S. Geological Surv, Professional Paper 14, p. 106. 



