ADJUSTMENT BY LEAST SQUARES 



177 



quently transformed into decimal fractions of a dyne. The excesses 

 varied in sign and size, many exceeding 100 units and the largest being 

 346. Many of the actual link-increments exceeded 1,000 units, and 

 as these depend on measured gravity-gradients of which the accuracy 

 rarely exceeds 2 to 3 per cent in practice, it is considered meticulous 

 to obtain the adjusted values of the link-increments to nearer than 

 10 units. The procedure thus resolves itself into one of obtaining from 

 the excesses the corresponding correlates to the nearest 10 units. 



TABLE VIII 



To facilitate setting down the related first- and second-remove cor- 

 relates corresponding to each key correlate, it is convenient to replace 

 every triangle in Figure 7 by a correspondingly numbered dot, situated 

 roughly at its center of position, and to join these dots by lines repre- 

 senting contacts of adjacent triangles. A triangle is thus represented 

 by a point from which radiates a number (one, two or three) of lines 

 to each of its first-remove points, from which again outwards radiate 

 lines to the second-remove points. Commencing at any one point it 

 is easy for the eye to pick up the first- and second-remove points in 

 turn. These are tabulated on the working sheets, of which Table IX 

 shows a typical example used for the first five points in order. The 

 first column, headed N, is used for the common divisors obtained 

 from formulae (9) and (10) by simply subtracting the number of 

 first-remove points from 9 in each case, since in this network n = ^ 

 and n^ = g throughout. 



The second column marked A contains the key-point number, fol- 

 lowed after an interval of one line by the numbers of its first-remove 

 points in order. The third column, marked -\-m' — , is used for com- 

 puting m' from the excesses given in Table VIII and equation (9) . The 

 excess for the key point is multiplied by 3 and set down opposite 

 that point; the excess for each first-remove point is set down directly; 

 the -f and — columns are totalled and the aggregate sum divided by 



837 



