TEMPERATURE CORRECTIONS. 



85 



ordinate in millimetres for a change of 1 C. From a mathematical standpoint the two seta of results 

 just mentioned are not independent, the one being deducible from the other by mean!) of the relationships 

 given in the first column of Table X. In reality, however, the two sets of results were to some extent 

 independent, because there were in all months days sometimes a good many days in which information 

 was lacking as to the mercury temperature, the temperature trace or the Vertical-Force curve, and the loss 

 was sometimes of one element, sometimes of another. The third column of Table X is in most months 

 based on both sets of results (i.e. on Vertical-Force and mercury-temperature changes as well as on 

 Vertical-Force and temperature-trace changes), allowing most weight to the direct Vertical-Force and 

 temperature-trace comparisons. The second and fourth columns in Table X are calculated from the 

 third column, employing the accepted scale values of the Vertical-Force curve and the temperature trace. 

 Any uncertainty in the scale value of the Vertical-Force curve thus affects these two columns, but it 

 does not and this is important have any influence on the accuracy of the temperature correction 

 actually applied to the readings of the Vertical-Force curve. The ordinates of the Vertical-Force curve 

 arid temperature trace were read in millimetres, and for each millimetre change of ordinate of the 

 temperature trace a corresponding correction in millimetres was applied to the Vertical-Force ordinate. 

 There is not the slightest doubt that the application of the temperature corrections immensely improved 

 the accuracy of the Vertical-Force results in fact, in some of the Midsummer months the results if 

 uncorrected for temperature would have been practically useless but considering the many changes of 

 scale value and other sources of uncertainty it would be too much to hope for complete success. 



19. The uncertainties remaining after the application of the temperature correction increase, of course, 

 with the probable error in the calculated value of the temperature coefficient, which there is no satisfactory 

 means of estimating, but they are equally dependent on the range of temperature of the Vertical-Force 

 magnet. So far as the regular diurnal inequality of V is concerned, it is only the regular diurnal change 

 of temperature that counts. It is the range of this regular diurnal change that appears in the first column 

 of Table XI. It is derived from the temperature trace on those days which were actually used in deducing 

 the diurnal inequalities of V in Table XVIII. 



TABLE XL 



