It is however necessary to cal culate a test function to the Empirical Photogrammetrie Distribution, Professor 
Bjerhammar in Stockholm indicated a method in his book "Geodesy"( Almqvist & Wiksell. Stockholm 1967, Page 285-286). 
2 
We calculated the Fmpirical Photogrammetrie X Distribution, Table 8,14, The F-distribution must be possible to use 
  
also on not normal distribution. (B, L. v, d. Waerden-Mathematische Statistik 1957, Page 247). 
9 
Application of the Empirical Photogrammetric X -Distribution. 
  
In one control point of a model the r, pz, p, or z values could deviate very much, from the correspon - 
ding values in other control points. 
The hypothesis that the deviating value could belong to the group of control points could be tested with this 
x distribution, 
Deviating values for one model could be compared similarly with the group of models. 
The variance for one model could similarly be compared with the population variance, stated as an error 
limit. 
A special problem for the future is to study how a few p values could be combined with many z values 
toa weighted r value. The weighted mean will be easily found, but the corresponding degrees of freedom 
could be difficult to state. In the meantime it is advisable to test only homogeneous data ofp , — z, orr. 
2. REGRESSION ANALYSIS 
9.1 The Regression of Point Height (p z) Errors of Comparison Points upon the Residual p z Errors or the 
  
i1ve ulven Control Points 
  
The program for multiple regression analysis was originally given by M. A. Efroymson in the book "Mathema- 
tical Methods for Digital Computers’. It has been corrected and also adapted to the computer Gier, used by the 
Swedish Cadastre and Land Survey Board. 
The regression function will be 
o 9 9 
f^: Asp. B 
The result of the regression was the following functions 
2 en” T aoi À + 2 
1:3 500 r 0.056- 0,037 4 2.464- 0,583)p + (0.964- 0.444)z 
2 + + 2 
1:6 000 r = 0°173- 0.246 +3, 509 - 0.326)p 
2 n + 2 + 9 
1:12 000 r zo 4.045- 0,914 - (0, 804 -: 0, 118)p +(0.792- 0.086)z 
In the Reichenbach experiment the result was 
o 
5 2 + 
ro i[1.141- 0.060)p^ *(1.013- 0.064) 
€ 
With more than 99 % confidence.