2 W. E. Alkins & J. Harwood— Variation of Sphceria 



Again, the correlation of each pair of measured axes has 



been studied, as has that of the two ratios = ? and 



Length 



Thickness 

 Length 



Results. 



(a) Length, Width, and Thickness Distribution. The 

 distribution of length, width, and thickness is shewn in 

 Table I (in all the distribution and correlation tables a class- 

 interval of 0*3 mm. has been adopted as most satisfactory). 

 The thickness distribution is very nearly symmetrical, and the 

 asymmetry becomes slightly more marked in the length and 

 width curves. The coefficients of variation for the three 

 dimensions are in very close agreement — length, 979; width, 

 10*81 ; and thickness, 1272. 



TABLE I. 



Distribution of Length, Width, and Thickness. 



Length 



No. of 



Width 



No. of 



Width 



No. of 



Thickness 



No. of 



mm. 



Shells 



mm. 



Shells 



mm. 



Shells 



mm. 



Shells 



3-9 



1 



5-3 



1 



8-6 



64 



24 



6 



42 



2 



5-6 



2 



8-9 



63 



2-7 



8 



4-5 



6 



5-9 



2 



9 2 



56 



3 



27 



4-8 



10 



6-2 



6 



9-5 



30 



3 3 



94 



51 



22 1 



6-5 



5 



9-8 



25 



3-6 



113 



5 4 



63 



6-8 



6 



10-1 



11 



3-9 



112 



57 



71 



7-1 



13 



J0-4 



7 



4 2 



93 



6-0 



117 



7'4 



24 



10-7 



2 



4-5 



34 



6 '3 



102 



7-8 



41 



110 



2 



4-8 



9 



6-6 



60 



8-0 



64 



11-3 



— 



5-1 



4 



69 



7'2 

 7'5 



28 



11 



5 



8-3 



73 



11-6 



3 





















7'8 



1 















8-1 



1 















(b) Correlation of Length and Width. The correlation 

 table for length and width is given in Table II. The co- 

 efficient of correlation has the value + 0*9628 ; the equations 

 of regression are : — 



(a) W = - 0*540 + 1*505 L, 



with the standard error ± 0*2488 ; 



(b) L = 0*773 + 0*616 W, 



with a standard error of ± o'i 592. 



The regressions are in all cases assumed to be linear. 



