;}(i DISCUSSION OF THE HORIZONTAL COMPONENT 



in order to give the correction to the means derived from five years, become, there 



tore: for months between July and December inclusive, +26; for January f 10; 

 for February, March, April, and May + 14; for June + 2. These corrections are 

 ( onstanl for each hour of the day in any one month, and consequently do not affect 

 the diurnal variation; but they have nevertheless been applied at once to facilitate 

 subsequent deductions. Their origin has also been explained in the remarks accom- 

 panying Table V. of the preceding part. 



The following example of the process of interpolation for the odd hour values 

 will suffice for all similar cases: Required the mean normal from the 5 year series 

 for ;">" 21 \ m A. M. in June (see tabular values and results below). The mean nor- 

 mals for the two last years at 4" 21i m , 5 h 21| m , and 6" 21| m , are 117(i, 117:3, and 

 1169 respectively; the mean at 5' 1 21 i"' is therefore 3 divisions less than the mean 

 at 4" 21*'", and since the mean of the ."> year series at 4" 21| m is 853, the result for 

 5" 2\\ m becomes 849; again, adding 4 divisions to 847, the mean at (i' 1 21 V". we 

 find 851; the mean of the two values, or 850, is that given in the table, to which 

 + 2 has been added, making the final result 852. The means of the odd hours, thus 

 found from the adjacent even hours, in general, do not differ by as much as a scale 

 division. 



The time given in the tables of the normals is mean local time, counting from 

 midnight to midnight to twenty-four hours. The observations were taken (on the 

 average) 21| minutes after the full hours, as indicated in the tables. Increase of 

 scale readings indicates decrease of horizontal force; the value of a scale division 

 equals 0.0000365 parts of the horizontal force, or 0.0001523 in absolute measure, 

 the mean horizontal force being 4.173 (in absolute measure). Proper weights have 

 been given to the normals of the even and odd hours, in proportion to the number 

 of observations, as will be seen hereafter. Other special remarks will be found at 

 the end of the month to which they refer. 



Tabulation of the hourly normals for each month and the mean of the five year 

 series, expressed in scale division readings anil reduced to tin 1 standard temperature 

 of 63° (Fahrenheit's scale), also corrected for all irregularities in the progressive 

 change. The regular progressive and secular change, therefore, remains in the 

 tabular quantities. 



