1879.] Report to the Committee on Solar Physics. 109 



Having grouped the whole 16 years' observations according to this- 

 L period, we next broke up these into yearly sets. Each of these sets 

 might thus be expected to be freed from the influence of the well- 

 known annual inequality of declination range. These yearly sets 

 embraced generally fifteen, but sometimes sixteen rows of twenty-four 

 each. Our starting-point or epoch in all these calculations was January 1, 

 1858, that is to say, in all our groupings the declination range for 

 January 1, 1858, formed the first member of a horizontal series. 



The next operation was to sum up these 15 or 16 sets for each of the 

 24 horizontal figures. It might naturally be supposed that we should 

 then divide each of the sums so obtained by 15 or 16, as the case might 

 be, and then find the difference of each of the 24 quotients from the 

 mean of all the quotients, such differences when placed together repre- 

 senting the inequality for that year. 



There appears to be, however, reason to believe that (leaving aside 

 all speculation as to causes) on those occasions when the daily range of 

 the declination magnet is greatest, the variations of this daily range 

 are greatest likewise, and possibly in nearly the same proportion (see 

 paper by Balfour Stewart on " The Variations of the Diurnal Range of 

 Magnetic Declination as recorded at the Prague Observatory," " Proc, 

 Roy. Soc," May 2, 1878). We have thus considered it as on the 

 whole safest and best to adopt this plan with respect to our yearly 

 results ; that is to say, we have regarded the mean of the 24 points for 

 each year as equal to 1,000, and we have represented each individual 

 point upon this proportional scale. 



8. The yearly inequality is, therefore, represented by the series made 

 up of the differences of each of the 24 points from this normal value. 

 As it is desirable in carrying out our programme to represent each 

 yearly inequality by means of a curve, we have to some extent smoothed 

 or equalised this series of differences. The primary series (A) has 

 been converted into another series (B), also of 24 points, each point 

 of (B) being the mean of four consecutive points of (A), and the 

 series (B) has then been converted by a similar process into a series 

 (C), each point of which is a mean of four consecutive points of (B). 

 This amount of equalisation, by getting rid of what we may term 

 accidental fluctuation is quite sufficient to enable us to draw a curve, 

 well representing each yearly inequality. The equalised yearly inequa- 

 lities corresponding to the period 24 - 25 days are represented in Table I. 



