202 KNOTT AND TANAKADATE 



If we take all the 81 stations, the mean probable error of a sin- 

 ofle observation comes out ± .00108. 



If we throw out Xos. 39, 41 and 73, which are again conspicuous 

 by the magnitude of their Üitferences, No. -Il ( Hakone ) being espe- 

 cially so, the i)robable error becomes ± .00091. It will be noticed 

 that No. 64, one of the Korean stations, is more conspicuous even 

 than No. 73 in res])ect to the nrignitude of the Difference ; and that 

 No. 19 (Nemuro) d(jes not fall far short of it. If these also be neglect- 

 ed, the probable error comes out ± .00084. 



If we take into account only the fifty selected stations, the 

 probable error is ± .00080. 



Here four Stn{i(ms stand out prominently })y virtue of their high 

 differences. These are Shiogama (No. 5), Höjö (No. 43) Ilagi (No. 

 72 ) and Hamada (No. 73 ). If we neglect these the probable error 

 becomes ± .00064. 



Two of these four neglected Stations, namely Hagi and Hamada 

 (Nos. 72 and 73 ), were also amongst the Stations that were similarly 

 treated in tiie discussion «ff the Dip. Kashiwazaki (No. 35), which 

 was one of the four neglected in the discussion of the Dip, is fairly 

 prominent also by reason of the magnitude of the diff'erence between 

 the observed and calculated Horizontal Forces. Hichiyamura (No. 

 68), however, the remaining one of the four neglected Dip observa- 

 tions, is characterised l)y an exceptionally small difference between the 

 observed and calculated values of the Horizontal Force. A similar 

 remark applies to Shiogama (No. 5) and Höjö (No. 43), whose 

 Horizontal Force differences are large, but whose Dip differences are 

 comparatively small. 



And now, expressing the co-ordinates in kilometres, we find 



H = .29482 — .00003335 9 - .00000785 X 



