13^ 



— ho= et cos X — ki, = « sin -/ ') 



tlie values of x were calculated for the separate years and these 

 values were united in 7 groups, as follows : 



Each group gives a normal value for x iind these are then repre- 

 sented by equations 



X = Xo + f* (^-1900.0). 

 To the first equation I gave the weight 0.7, to the others the 

 weight 1. These being solved by least squares gave the result 

 X„ = 1107°.! = 27°J fi = 21°.085 

 The last column of the table gives the differences between obser- 

 vation and calculation. The annual variation of the argument found 

 is thus 0°.43 larger than that which follows from the theory, 20°.65. 



Newcomb found 21°. 6 

 Bakhuyzem ,, 19°. 36. 



The argument for 1900.0 is now found M°. 3 larger than according 

 to the theory ; for the mean epoch of the observations 1886 the 

 difference, however, is only -\- 8°. 3, while my previous calculation 

 mentioned above gave Obs. — Th. = - — 2°.0. 



We now proceed to the investigation of the residual values for 

 — /zy and — ki; which after correction for all BroWxN's terms, still 

 show a distinct periodicity, though the amplitude is greatly decreased. 



By a graphic representation and some preliminary calculations I 

 came to the conclusion that the best agreement would be attained 

 by a term of a period of nine years. The values of h and k seemed 

 to agree completely in this and together to point to the existence of 

 a term of the form « sin [g -f x)- 



1) Our jc is connected with the N introduced by Newcomb and also used by 

 E. F. V. D. Sande Bakhuyzen by z~ ^ — ^^'• 



For he and kc the values were taken according to the 2nd calculation. 



