34 A. TANAKADATE. 



iVoni wliieli we see tliat tlie quadratic formiilre are sufficiently 

 near for the purpose. Some of the coefficients whose values are 

 less than probable errors may be omitted; but since the omission 

 of only one or two terms little lightens the labour of com- 

 putation, they are all kept for the sake of uniformity. 



The reduced values of mapiietic elements in 241* stations 

 which were seemingly free from large local disturbances w^ere 

 treated in the same manner and gave 



r,= .5° 03:i5-8.'274J;. + 17:36r)J^-0:649J7^-0^23GJlJ^-0:075J^' 

 + .68 ±.291 ±.34.'5 +.09G ±.187 ±.103 



^ = 50° .50.'G1 -7:578 j;.+ G8:253J^ + 0.'296JÎ^-0'43SJIJ^-- 0.482 J^- 

 ±.58 ±.247 ±.292 ±-081 ±.158 ±.087 



H=2940L4-74!97J;.-362!4.5Jç^ + 3l497J^--l!316JlJ7'-4.331Jp 

 ±8.1 ±3.45 ±4.10 ±1.141 ±2.216 ±1.222 



wdiere j;.=(>^-138°)° and J^ = (^— 37°)° expressed in degrees. 



Expressed in minutes tliey l^ecome 

 0= 5° 03a5-0:i379Jyî + 0:2894Jçr-0:0001803Jl2 



- 0:00006.57 JlJ^- 0:0000209 J^- 

 ^=50° .50:61-0^263//;+ 1. '1376 J^ + 0.'0000821J^- 



- o:oooi2i8 j;j^-o:oooi340J^^ 



H =29401^4- 1.2494J;. - 6!o409J^- + 0^0009713 Jl- 



- 0^0003656 j;j^ -0.0012032J^c^' 



The computation was carried in dujilicate beside being con- 

 trolled in the usual way. The mean probable error of a single 

 observation is 



* These stations are distinsuislied bv non-braeketed nniul)ei's in Tables XV and XVI. 



