FROM ORDINARY DAYS OF THE ELEVEN YEARS 1890 TO 1900. 
237 
§ 16 Values having been obtained from the absolute observations of each month 
corrected for diurnal variation, the means for all the months of the same name were 
summed and meaned, the 11-year period and the later period being treated 
separately. The monthly means thus obtained represented the average annual change 
from mid-January to mid-December, comprising the annual inequality and the secular 
change. It remained to eliminate the secular change. 
In the case of the 11-year period, the mean annual secular change of H between 
1890'5 and J900‘5 derived from the quiet days was +25'9y. A small difference has 
been observed at several stations between mean annual values of H derived respectively 
from quiet days and from all or all ordinary days, the former mean exceeding the 
latter. The difference, however, is only of the order 3y, and as 1890 and 1900 were 
both quiet years of very similar character, any uncertainty of this kind spread over a 
10-year period must have been negligible. A confirmation of the accuracy of the 
quiet day estimate was derived by taking arithmetic means for 1890 and 1900 of the 
twelve monthly means of the absolute observations, corrected for diurnal variation. 
The mean thus obtained for 1900 exceeded that obtained for 1890 by 259'5y, giving 
+ 25'9 5 for the mean secular change. 
In the case of the second group of years, the mean secular change in II—as obtained 
from the annually published quiet day results, allowing for changes of constants and 
procedure—was only +5'8 2 y. In the case of I the mean secular changes accepted 
were — 2 /- 25 5 for the first period and — l r '05 5 for the second. 
The substitution of these respective values for AM in the formula in § 15 led to the 
annual inequalities for H and I given in Table XVIII. The annual inequalities given 
for V in that table were calculated from the formula expressing changes in V in terms 
of changes in H and I, employing mean values for the numerical coefficients of AH 
and Al. At the foot means are given for the ordinary three seasons. The centre of 
each season falls at the middle of the year, so these seasonal values are unaffected by 
the secular change, or by any error that may have been made in estimating its 
amount. 
The greater or less smoothness of the inequalities, and the amount of accordance 
between the results from the two periods, are the chief criteria for estimating the 
reliability of the data in Table XVIII. Both criteria are less favourable towards the 
inequality in V than towards those in TI and I. The V inequality from the 11-year 
period is very irregular, plus and minus signs occurring rather promiscuously; and 
while the corresponding seasonal means from the two periods agree in sign, the winter 
and summer means from the second period are numerically much larger than those 
from the first. The lesser consistency of the V inequality is hardly surprising since 
it suffers from every uncertainty or accident that affects either the H or the I 
inequality. A change of lOy in H alters I as much as a change of 24y in V. Thus 
the deduction of changes in V from the combination of observed changes in I and H, 
though the only way feasible, does not promise high accuracy. 
2 k 2 
