BETWEEN SUN-SPOT FREQUENCY AND TERRESTRIAL MAGNETISM. 159 



implies a very conspicuous difference between the phenomena in years prior and 

 subsequent to 1864. Not improbably the want of homogeneousness in the earlier 

 data, already referred to, may be partly accountable for the apparent change. At all 

 events, the results for the final period, 1889-96, show no distinct progressive 

 diminution of a or increase of I as compared to the period 1865-96. In fact, the 

 " all " day data for these two periods, and the " quiet " day data for the shorter 

 period, give almost identical values for the mean b for the year. 



As regards the seasonal phenomena, each set of Greenwich data makes b con- 

 spicuously least, but b/a conspicuously largest, in " winter." The December value of 

 b is invariably the smallest. The "all" and "quiet" day data for 1889-96, as 

 already stated, give very nearly the same mean value of b for the year, but the 

 " quiet " days' value for b is much the larger of the two in winter, and the smaller in 

 summer. The ' ' quiet " day data at Greenwich present a remarkable similarity to 

 the corresponding Kew data, as is best seen by comparing the seasonal values for b/a. 

 The fact, however, that the absolute values of both a and b are some 10 per cent. 

 higher at Greenwich than at Kew is rather suggestive of some misapprehension as to 

 the scale values at one or both observatories. 



9. In cases such as the present, a comparison of calculated and observed values is 

 useful. The Greenwich data do not lend themselves very readily to this, as 

 Mr. ELLIS does not give mean values for individual years, and the interest attaching 

 to such a comparison for individual months of the year seems hardly sufficient to 

 justify the necessary labour. In (A), 75, some results were given for individual 

 months of the year at Kew, but none for the year as a whole. This information is 

 accordingly now supplied in Table IV., so far as concerns the ranges and the sum of 

 the 24 hourly differences from the mean in the mean diurnal inequalities for the year 

 in the several elements. The calculated values are derived from the values given for 

 o and b in (A) Table XLIV. From the mathematical standpoint the nicety of 

 agreement is best judged by considering the last line in Table IV., which shows what 

 percentage the probable error in a calculated value is of the total variation exhibited 

 by the element in the 11-year period. For practical purposes the mean difference 

 between the calculated and observed values, and the percentage it forms of the 

 absolute mean value of the element, are, however, fully as important. 



It will be seen that the agreement is about equally good for the ranges and for the 

 sum of the 24 differences. It is, on the whole, slightly better for D and H than for 

 V and I. 



