11 
dy 1 , " , 
wg dx, — 3 dx, — 2 dx, + 3 dx; — 2 dy, — 3 dy, + 
y , LA , " , 
"fI (2 day — 3 dx, + 3 dæ, — 2 dx, — 2 dy, + 2 dys) + 
rg (da, 4- dz, — da; — dz, — dy, + dys)j — 
y d , d " x d , x 
Tuy red Ard = —1 dy (29) 
As emphazised above, the differentials da’ etc. are not direct image 
coordinate measuring errors in a strict sense. As demonstrated by 
HALLERT 1957 b the differentials da’ etc. can be written as functions of 
differentials du’ dv’ ete. which are to be regarded as real image coordinate 
measuring errors 
, y , , 
dx’ = du’ — du, + d (dv, — dv) (30) 
1 1 lv, : 1 T 
dy =dv —do.{1 ——)—dv, + 31 
y dv dv, b dv, (31) 
" y ” " 
dx" = du” — du, + S. (dv, — dvi) (32) 
; 2 
ly" — dv" —d t dendi =: 3 
" == Vin y" 1 Lie im 8 3: 
dy dà dv, b dv, (33) 
The weight and correlation numbers of these expressions were de- 
monstrated above in the expressions (5)— (7). 
For a theoretically correct investigation of the error propagation and 
distribution the next step should be to apply the general law of error 
propagation to the expressions (28)—(29) in combination with the 
(33) as demonstrated in 
  
weight- and correlation numbers from (30) 
(5)—(7). This procedure would. however, become very laborious. There- 
fore we shall treat the error propagation as though the differentials 
dx' ete. were errors of direct image coordinate measurements. The 
discrepancy introduced by this approximation and simplification is 
negligeable and can be overlooked.