About 85 % of the pairs of coordinates of a comparison point had no partial correlation. Table 6.23. Inthe 
  
   
Experiment Reichenbach 1962-1964 no partial correlation was found. This indicates that the (x y z) errors V 
does 
of a comparison point are independent of each other, so that a large x error not indicate a large z error. 
  
  
  
  
  
  
  
  
  
P 
Most tests in this Experiment indicated no significant difference between inner and margin points o f a model. t 
Table 6.24. In the Experiment Reichenbach 1962-1964 there were small differences between these two groups 1 
of points. In practice consequently these two groups of points are considered to have equivalent accuracy. V 
e 
Section 7. Error of Distance. The standard deviation and correlation coefficients for classes of distances t 
were calculated from differences 
‘between geodetically and photogrammetrically determined distances. Table 7.1 and Figures 7.11, 12,13. a 
The corresponding normalized curves are presented in Figure 7.14. We found the standard deviation increase : m 
with length £r distances shorter than one base length, be constant for distances between one and two base V a r 
lengths and decrease with length for distances longer than two base lengths, C 
t 
Section 8. Empirical Error Distribution. E 
Already in the Experiment Reichenbach 1962-1964 we found the photogrammetric error distribution beeing r 
not normal. We presented these distributions in some tables, together with theoretical test functions. In the 7 
Experiment Pecny 1966-1968 we calculated the absolute frequence and cumulative pz distributions and pre- 8 
sented them graphically. Figure 8.1 and 8. 2. The distributions are scew in z with more + errors than - errors, 
It is namely possible to observe stereoscopically that a measuring mark is above a surface but not if it is under 1 
the surface. 8 
The relative frequencies were also calculated from all negative scales grouped together as The Empirical f 
Frequency pz Error Distribution, Table 8.11 and The Empirical Photogrammetric Cumulative pz Error Distri - 1 
bution Table 8.12. The maximum pz errors of these distributions were found with the aid of a sorting pro- é 4 1 
cedure using a computer.They are presented in Table 8. 13. Only the z errors are nearly as large as but : 1 
still smaller than some criterias in section 5. for excluding gross errors. C 
The Empirical Photogrammetrie x" Distribution was calculated and is presented in Table 8. 14. 4 
We simply calculated all possible 5 Fre 2 , where 2” is one variance of all available variances and M is 
the mean variance. We sorted them in classes of magnitude and so we had the X distribution for ane degree ] 
of freedom. Then we calculated all possible EN and sorted them, after that all possible V 
Gs etc. up to 15 degrees of freedom. The number of quotients in each class was the absolute S 
frequencey. Corresponding relative frequency gave the probability. 
Section 9, Regression Analysis. 
  
Because of lack of time the summary and conclusion will be presented at the Congress separately.