6. CONCLUSIONS 
Before drawing conclusions, first some general remarks have to be made 
- Looking to the boundary values (tables 3 to 6) one sees that in all 
cases the values for Vx and Vy are the same. It is easy to prove 
that, if the a priori standard deviation for the x- and y-model 
coordinates are equal and one choses a variance covariance matrix 
bases on the linear function 3.1 (which results in circular 
standard ellipses), Vx and Vy are indead also equal. For the sake of 
completeness both Vx and Vy are printed. 
- Looking to tabie 2, the most right column shows a couple of times the 
value 2.0. This is always the case, where it is concerning control 
points. After the least squares adjustment, the variance covariance 
matrix (as part of the whole v.c. matrix) is equal to that which is 
derived from the earlier mentioned linear function with parameter 
c,=2. Now in the eigen value problem the criterion matrix H is 
derived from exactly the same function, only with c,-l. 
When in the eigen value problem, a partial network is chosen, which 
only consists of control points, it can be proved that, calculating 
the eigen values after the second phase of the adjustment, all eigen 
values are equal.to 2.000. 
- In tables 3, 4 and 5 not all boundary values of all control points 
are printed because of reasons of symmetry of the associated blocks. 
Evaluating the results the following conclusions can be drawn. 
1. The precision of each block is decreasing significantly when 
connecting it to control points. This is mainly due to the fact that 
the control points, which are situated at the periphery of the block, 
have a great impact on the variance covariance matrix of the 
associated tie-points. These tie-points, especially those which lie 
in the corners of the block (compare partial network lb in the first 
experiment) determin for a great deal the big values of ax in the 
first phase. | ; 
Remark: Since the early seventies a lot of experience is obtained 
with judging the precision of networks. À good precision is obtained 
if the progress of the maximum eigen values is about linear with 
respect to the number of points of the considered partial networks. 
2. The maximum eigenvalue of the second phase increases when dealing 
with more tie-points (compare experiments | and 2), whereas this 
effect is overruled in the first phase by the bad precision of the 
tie-points on the periphery of the block. The effect on the 
reliability (compare tables 3 and 4) is negligible; 
3. The first phase precision of a square block is better than that of 
a rectangular block, but this effect is eliminated after the 
pseudo least squares adjustment (compare eigenvalues of experiments 
4 and 5). 
4. The impact of extra tie-points through the middle of the block on 
precision as well as reliability is negligible (compare 
experiments 4 and 5).