NO. 6.] 



INTRODUCTION. CHRONOMETERS. 



LI 



This material was examined on the supposition that the rate could be 

 represented in the form 



Daily rate = x -f- ty 



where t is the temperature. For Htv after the return an attempt was also 

 made to introduce a term proportional to the time, but it was found quite 

 insensible. The results are contained in the following synopsis, where n is 

 the number of equations employed. 



It will be seen that Kt has the smallest temperature-coefficient, but that 

 its constant term has changed considerably more from 1893 to 1896 than the 

 constant term for the other two. From Table d it is also apparent that a 

 similar change in the opposite direction has taken place in the interval. The 

 relatively large probable errors of the constant terms depend chiefly on the 

 choice of as the standard temperature, the mean temperature during the 

 comparisons being of course considerably higher. 1 The last column gives 

 the mean of the squares of residuals (v) as a good means for comparing the 

 qualities of the three chronometers. 



Hohwti is evidently the best of the three and it was deemed safest to 

 rely solely upon it for the intervals without observations. A formula deduced 

 from comparisons ashore is of course not immediately applicable on board, 

 because the exterior conditions, especially the humidity, in a narrow ship's 

 cabin are very different from those of an observatory room. That these 

 different conditions affect not only the constant term but also the temperature 



' For Kt in 1893 it was found more convenient to count the temperatures from 10, 

 and the constant term was found as O.s32 0. S 194 for this temperature, so that 

 the probable error for should have been considerably higher than that given 

 above; but as the mean temperature during the comparisons in Hamburg had been 

 higher than in Christiania, the value for 10 was retained in order not to prejudice 

 the chronometer as compared with the other two. 



