( 404 ) 



§ 15. Representation of tlw oh.^ii'rr itions hy means of a four 

 term formula. 



We have now quite carried out the calculation of a formula of 

 the form 



' ^ A / ? ^' / ^ \' r t 



E—a 



100 



+ ^> 



100 



+ '• 



too 



+ 



100 



{C) 



announced in note 2 of § 11, by the method of E. F. v. d. Sande 

 Bakhuyzen, which proved to facilitate matters greatly again. 



Four solutions (C) were found, viz. (CI), (CII), (CllI) representing 

 the observations down to — 253°, whereas in (CIA ) only agreement 

 down to — 217° has been sought for. 



The coefficients in millivolts are the following : 



The residuals have been given in tenth parts of microvolts in 

 Table IX under W—Rci, W—Rcii, W—Rcjii, W—Rciv- 



Just as with the five term formula, the residual at — 182° 

 appeared also now greater than the others. 



In calculation 3 it was tried to distribute the errors more equally, 

 but the sum of squares appeared now to have increased. 



The mean errors are if we include the observations down to 

 — 253° for (CI), (CII), (CIII), and only those down to —217" for 

 (CIV), for 



(CI) ± 3.0 



(CII) ± 2.9 



(CIII) ± 3.0 



(CIV) ± 2.3 



If — 182° is excluded, they become: 



(CI) ± 2.7 



(CII) ± 2.6 



(CIV) ± 1.8 



The mean errors of (CI), (CII), (CIII) must be compared with 

 those of (^I) and (^III), ihose of (CIV) with those of {BIY). 



