236 REPORT—1896. 
% 
§ 8. Table V. shows how much more important relatively the non- 
cyclic effect is in the winter than in the summer half-year. 
In the winter half-year we see that the non-cyclic effect in both 
horizontal force and inclination is equal to about one-fifth of the range. 
This does not of course imply that there is an uncertainty of 20 per cent. 
in the range, because, whatever be the nature of the correction applied to 
eliminate the non-cyclic effect, it is hardly likely to introduce more than 
a small fraction of the observed difference between 0 and twenty-four 
hours into the algebraic difference of the maximum and minimum read- 
ings. The interval of time between these readings is in most cases nearer 
six hours than twelve. The fact, however, remains that in some indi- 
vidual winters the uncertainty as to the range must be very appreciable. 
When we come to individual winter months, notably January, when the 
observed range is least, the uncertainty is apt to be considerable. 
The preceding remarks refer exclusively to the uncertainties which 
the existence of the non-cyclic effect introduces into diurnal ranges 
deduced from ‘quiet’ days. Previous reports of the Committee! have 
dealt with differences between the ranges deduced from unrestricted and 
from ‘quiet’ days. It seems to me, however, that such comparisons are 
open to criticism so long as the proper treatment of the non-cyclic effect 
remains uncertain. 
Relation of Non-cyclic Effects to Diurnal Inequalities, 
§ 9. In the yearly and half-yearly results the most critical point is the 
nature of the diurnal inequality in the late night and early morning 
hours. The observed variation is then small, especially in winter, so that 
a disturbing element of no great absolute magnitude might completely 
alter the character of the phenomena. This will appear at once on refer- 
ence to the curves of declination and horizontal force in last year’s 
report, pp. 212 and 220. The curves on p. 220 are certainly suggestive of 
the presence of some abnormal influence during the midnight | hours ; at 
the same time this is not more true of them fhe of curves of the sainig 
type for Greenwich which Sir G. B. Airy ® based on data derived from all 
days but those of considerable disturbance. 
Elimination of Non-cyclic Effect. 
§ 10. If diurnal inequalities are to be got out at all from ‘quiet days’ in 
a form suitable for harmonic analysis, they must be made cyclic, and there 
is certainly no simpler way of doing this than that adopted last year, viz., 
treating the observed data as if the non-cyclic effect proceeded uniformly 
throughout the twenty-four hours. This method of treatment doesnot 
prejudice the facts. Supposing the non-cyclic effect to proceed irregularly 
throughout the twenty-four hours, then it may most conveniently be ana- 
lysed into terms, one being a linear function of the time, the others periodic 
functions whose periods are twenty-four hours or submultiples thereof. 
The linear term is eliminated by the method adopted last year. The 
cyclic terms of course remain, and are incorporated with the other cyclic 
terms of like period which go to make up the diurnal inequality on ‘ quiet’ 
1 B.A. Report, 1886, p. 71. See also paper by Messrs. Robson and Smith, 
Phil. Mag., August 1890, p. 142. 
2 Phil. Trans. for 1863 and for 1885. 
