( 249) 



The second members are written in 2 parts. The first part is 

 derived from the observed rates (I, the second from the prelimi- 

 narily rednced compensated rates Qi^ . The solution of the two sets 

 yielded : 



.V = -f 3.90 — 0.04 = + 3.86 



// z=z — 2.30 — 0.34 = — 2.64 



z = -\- 2.72 4- 0.46 = -f 3.18 



«= — 2.86 — 0.09 = — 2.95 



According to these results the influence on the daily clock lales 

 of the atmospheric pressure per 1 mm. mercury of 0° C. is -|- 0^0154, 

 of the temperature per 1 centigrade (upper thermometer) — 0.0264, 

 and of the difference in temperature per 1 centigrade 



(upper— lower) -f 0,318, 



while for the personal difference Bakhuyzen— Pannekokk in the clock 

 corrections is found — 0^.220. 



Table I contains the values required for the said comi)ntations; 

 for each interval between two consecutive determinations of the clock 

 correction : the number of days n, the observed clock i-ate Q, the values 

 Q^ , Q^, Q^ and Qf" , and also the values ((V, — Cr)'^ , {c,,—rrf, [c^—Cr) ^'' , 

 {c%j — tv)^^. To this must be added in order to render the compulation 

 possible the values for the last epoch Jan. 14, 1904, : 



!/^ = — 14.0 if = -f 57.7 gV — _s.l !/P — + 2.4 

 cB — _|_ 1.09 c^ — + 2.01 c^ — ^ 0.76 cP + 0.54 



</ = — 146, adopted in case we use the series of the observed clock rates Q. 



(jT ^^ — 194, ,, ,, ,, ,, „ preliminarily reduced Q/^. 



§ 7. Here follows the reduction of the relations between the mö^t 

 probable errors and the determination of the latter according to these 

 relations. 



Sub § 5 1 have derived for each error /i^ the relation : 



^II f o öi/, 1 



I ha\e thought myself justified in adopting llie same value for the 

 mean errors in the determinations of the clock correction, although 

 for 1 of fhose corrections only 2 stars, for 13 corrections 4 stars, 

 and for all the others 3 sfars were observed, since the difference in 

 accuracy resulting from the different numbers of stars is relatively 



