errors 
d by 
ed as 
ut to- 
ee of 
, and 
syste- 
oeffi- 
cient 
1-test 
effi- 
ween 
per 
Table 10. The test of systematic errors within a set of models with the function z on different levels L 
of significance in per cent. 
  
  
  
  
  
  
  
  
  
  
  
  
Core Strip I Strip II Strip IIT 
Models : 
dinates zZ | e Z £ 2 £ 
X 1.854 — 2.116 5 4.923 0.01 
A Y 3.050 1 5.990 0.01 5.989 0.01 
A 4.890 0.01 0.019 — 4.556 0.01 
X 4.735 0.01 0.885 — 7.444 0.01 
B Y 3.319 0.1 1.474 — 2.978 1 
Z 3.556 0.1 1.420 — 4.181 0.01 
  
  
The conclusion might be that there are systematic errors within a set of models for the strip I and III. 
However no conclusion could be made as to the strip Il. 
6.33 The magnitude of the systematic errors for each of the models might be estimated with the aid of 
S?,. In the following table the weighted, quadratic means of S, are given for the strips. 
Table. 11. The weighted quadratie means of S,. 
  
  
  
| ol S, in mm for 
Strip 
X | Y | Z 
I | 64 86 | od 1 
II 77 84 134 | 
111 | 126 114 | 149 
  
6.4 The random errors. 
6.40 The random errors S9 of every model were given in the tables 1: I. 1. ll] as to X, Y, Z. Also the 
errors S, of every model were given in the same tables as to X, Y, Z, LA, p and L4. p. 
6.41 The hypothesis that the variances S?, within a set of models would belong to the same population 
was tested. (Dixon & Massey. Introduction to Statistical Analysis. Page 90 and Biometrica Tables for Sta- 
tisticians I. Page 57.) The result of this test of heterogeneity at the 0.1% level of significance is given in 
the following table. 
Table 12. The test of heterogeneity for variances within a set of models (0.1 % level 
of significance) 
  
  
  
  
  
  
  
  
  
  
  
  
Test without "m : 
est, wituO l'est after that some models were removed 
Lali removing of models 
Strip 
X | Y | Z Removed X | Removed Y | Removed Z 
I Heterogeneous 1A, 20A Homog. | 1A, 204, | Homog. | 1A, 204, | Homog. 
91A 91A 91A 
IT Homog. | Homog. | Heterog. Homog. Homog. Heterog. 
JIT Heterogeneous Heterog. Heterog. Heterog. 
  
  
. . (9 
In the same way was also tested the hypothesis, that the variances 5^, would belong to the same popula- 
tion. The result was that this could scarcely be the case at the 0.1 ?/o level of significance. 
Therefore the conclusion might be, that the F-test and similar tests could not be used for testing the signi- 
ficance of differences between strips concerning X, Y and Z. Nor could such tests be made for testing diffe- 
rences between X and Y in the same strip. 
6.5 The total or the combined error 
6.50 It was showed in this paper, that the constant errors within single models must be considered as 
random errors within a set of models (Compare 6.2). However it was hardly made clear, that the syste- 
matic errors within single models could be considered as random errors within a set of models. (Com- 
pare 6.3). The random errors within single models could hardly be comparable from statistical point of 
view. Nevertheless such measurements have been regarded as comparable in the photogrammetric prac- 
tice. (Compare 6.4). 
15