Table. 2. The coefficients ai, /3y ,■ awd Sj* 
Point 
a\ 
Bi 
y. 
/ 1 
Si 
1 
+ 0.441 
+ 0.555 
— 0.441 
4- 0.293 
2 
+ 0.236 
0.000 
0.000 
4- 0.246 
3 
+ 0.441 
— 0.555 
4- 0.441 
4- 0.293 
4 
+ 0.058 
+ 0.511 
— 0.498 
4- 0.178 
5 
— 0.193 
0.000 
0.000 
4- 0.086 
6 
+ 0.058 
— 0.511 
4- 0.498 
4- 0.178 
7 
— 0.209 
+ 0.729 
— 0.811 
— 0.341 
8 
— 0.622 
0.000 
0.000 
— 0.591 
9 
— 0.209 
— 0.729 
4- 0.811 
— 0.341 
adjusted, The problem can be solved by means of statistics. Prelimi 
narily the level of 33.3 % has been used |3| and the adjustment has 
been performed when dt and dl havre exceeded their standard errors. 
With the assistance of different organizations several A8 autographs 
have been tested and adjusted according to the method in question. 
The results are given in the following tables No 3 and 4. 
Table 4 shows clearly that the width-error is highly significant 
before the adjustment. The autographs have been adjusted according 
to the usual instructions by the manufacturer [4]. Thus unsymmetrical 
locations have been used in the x-direction on account of which the 
possible terror can be adjusted with the width-error. 'Phis incomplete 
determination of the width-error must be considered evidently unfit 
for use. 
After the adjustment 83 % of the errors is on the level of 100.0— 
33.3 %, 12.5 c /c of the errors is on the level of 33.3—5.0 %, and 
4.2 % of the errors is on the level of 5.0—1.0 %. Consequently it 
seems to be suitable to have the level of 5 % as the limit for the 
adjusting. 
4. The problem of the control. 
To know how frequently the adjustment situation of the autograph 
must be controlled is of great interest. Certainly this depends upon the 
time during which the x-obliquity and the width-error are under the 
confidence limits. Therefore an investigation in the autograph A8— 
721 has been performed where control measurements were made during 
2.5 years. The results are given in the following tables No. 5 and 6. 
447