NIPHER — MAGNETIC DETERMINA lIOxNS IN MISSOURI, 79. I 29 



Observed time of lOO oscillations, 



Time of one oscillation, 



Correction for rate, 



T' 



The correction for the rate of the watch is one-fifth of the cor- 

 rection for a change of temperature of i° F., and hence has been 

 omitted in the reduction of all the work. 



The watch was compared several times during the summer, 

 and was found to run very uniformly, losing 3' per day. It was 

 allowed to run down two days before returning to St Louis, and 

 hence no comparison could be made at the close of the work. 

 When lying face upward, as in making observations, it was found 

 to gain 4^^ per day. 



The temperature corrections and moment of inertia as function 

 of temperature have been used in the present reduction as deter- 

 mined in 1S78. (See pp. 87 & 90.) 



Two deflection series were made at each station, the distances 

 (r) between the deflecting and the suspended magnets being i.^^ 

 ft. and 2. 00 ft. In the latitude of Iowa City, the suspended mag- 

 net was deflected to the limit of its scale when the two magnets 

 were 1.75 ft. apart. The deflection bar was only 4 ft. in length, 

 so that while the values of r are perhaps not the best, they were 

 the best that could be chosen under the circumstances. 



The values P deduced from the deflection series of each station 

 are given in a table following (see next page). 



The formula for the calculation is given in the former report, 

 p. 90. 



The indiscriminate mean of these values {P) is — o.^^^^^' 



By the application of Peirce's criterion, the values for July 5th 

 and September 13th (the first and last of the series) are rejected : 

 hence the final value of P becomes 



^=+O-0105^ 



Probable error of mean, rb o.^Qg^. 



r P^ 



The special values of the factor i _ ^ become 



log- [ -I (^^ J =. 9-99887 



log. 



ri_^^ 



