88 
ME. C. CHAMBEES ON THE DIEECTION AND INTENSITY 
February, about the winter solstice. The difference of range of the four-monthly means, 
imcorrected and corrected, is doubtless a consequence of the uncorrected values having 
reference to the mean interval from 12 h 22 m to 1 4 h 39 m , whilst the corrected values have 
reference to the mean of the twenty-four hours. 
Whilst, however, recording this result as that which the absolute observations are 
capable of yielding, the writer does not claim for it any great confidence ; for the pro- 
bable error of an observation is considerably greater than the monthly differences which 
have been obtained as representing the annual variation ; and consequently it is only by 
combining together the observations of a number of years that the influence of errors of 
observation can be expected to be practically eliminated, and half the period of six and 
a half years seems not to be sufficient for this purpose ; for when the observations are 
separated into two groups (of three and a half years and three years) the annual varia- 
tions yielded are of totally different character. 
19. Probable Error . — The probable error of a single weekly determination of Absolute 
Horizontal Force has been computed separately for July to December 1867 and for the 
year 1868. This has been done independently from all the complete observations except 
those of July 2 to 27, 1867, and for all, with the further exception of those of August 
21, 1867, and April 8 and 15 and August 19, 1868, which give results that are 
evidently erroneous, far beyond the range of observational error. The observations were 
all reduced to the constant reading 24‘27, at temperature 82°, of the large Horizontal- 
force Magnetometer, allowance being further made for the loss of strength of the 
magnet of that instrument ; and the differences being then taken between each corrected 
determination and the mean of all, the probable errors were calculated from these 
differences by the method of least squares. The rate at which allowance was made 
for the loss of strength of the magnet was "00015 of the whole per annum (see ‘ Intro- 
duction to Bombay Observations,’ 1864, page xvi). The probable errors found are as 
follows : — 
Table XIX. 
Period. 
Including abnormal values. 
Excluding abnormal values. 
English Units of Force. 
Of a single 
weekly 
determination. 
Of the mean of 
all the 
determinations. 
Of a single 
weekly 
determination. 
Of the mean of 
all the 
determinations. 
July to December 1867 
January to December 1868 
+•0053 
+ ■0067 
+•0011 
+ ■0009 
+ •0043 
+ 0043 
+ ■0009 
+ 0006 
The remaining observations of 1869 to 1873 being of about the same quality as those 
of 1868, will have about the same probable error. 
The maximum probable error (that of the moment of inertia of the vibration-magnet) 
