236 DR. C. CHREE: DISCUSSION OF KEW MAGNETIC DATA 
observing conditions are more or less favourable. Still in most months the mean 
times of the vibration and deflection experiments fell so nearly at fixed hours that 
one might have applied the same correction to each observation, basing it on a mean 
time from all the observations. As a matter of fact, however, this simplification was 
not adopted when dealing with the 11 years 1890 to 1900. After 1900, as already 
explained, magnetic conditions were less favourable at Kew. Still electric tram 
disturbances are, on the whole, of a nature calculated to impair the accuracy of 
individual observations rather than that of the arithmetic mean of a number of 
observations. It thus appeared worth while investigating the annual inequality 
deducible from absolute observa/tions made in years subsequent to 1900. 
As corresponding diurnal inequalities from ordinary days were not available, the 
corrections applied to the absolute observation results to bring them to the mean value 
for the day were based on the inequalities of the years 1890 to 1900. In the case 
of H, the years 1901, 1902, 1903, 1910, 1911, 1912, and 1913, having a mean sunspot 
frequency of 8‘8, had corrections applied from Table VII.,-which is based on the years 
1890, 1899, and 1900, with a mean sunspot frequency of 9’6. The years 1905, 1906, 
and 1907, having a mean sunspot frequency of 59'9, had corrections applied from 
Table VI., which is based on the years 1892 to 1895, with a mean sunspot frequency 
of 75'0. The two years 1904 and 1909, having sunspot frequencies of 42'0 and 43'9 
respectively, had corrections applied from Table V. for the 11 years 1890 to 1900, with 
a mean sunspot frequency of 41'7. The year 1908 was omitted, because the deflection 
distances were increased from two to three at midsummer, and this possibly might 
have introduced some discontinuity. The final outcome was that the twelve years 
dealt with had corrections applied as if their mean sunspot frequency were 31, whereas 
it was really 27. As we shall see later, the amplitude of the diurnal inequality in H 
and the sunspot frequency are connected, at least approximately, by a linear relation¬ 
ship. Extrapolation from one period of years to another is of course always a matter 
of some uncertainty; still there is considerable ground for believing that the accuracy 
of the inequality derived from the second period of years is not greatly inferior to that 
derived from the first. 
In the case of I the observation had been taken almost invariably in the afternoon, 
the mean time of observation falling within 30 minutes of 3 p.m. Near this hour 
I changes slowly, and its departure from the mean value for the day is not large. To 
have got out diurnal inequalities of I for each month of the 11 years would have 
entailed an immense amount of labour. Thus corrections to the observed values were 
simply derived from the mean inequalities from the 11 years given in Table XIV., and 
the corrections were calculated for the mean time of all the observations of the month, 
not for the times of the, individual observations, unless these times varied more than 
was generally the case. The I observational data from the 14 years 1901 to 
1914 were similarly treated, use being again made of the diurnal inequalities in 
Table XIV. 
