8 
For example for m = 4 and n = oo the 5 percent value of F is found to be 
s o 
2.3 7. This means that the ratio — becomes 1.54 and consequently s 0 should 
s 0 
not exceed 1.5 4 s 0 . 
It is of fundamental importance to determine the relevant value of s 0 to be 
used in instrument tests. The natural source of this information should be the 
manufacturer of the instrument, who knows his product. 
If such basic information is not available from the manufacturer, it can be 
obtained by a sufficient number of tests of representative instruments. Inter 
national test measurements within the ISP would be appropriate for this pur 
pose. Reference is made to the report (see HaJlert-Ottoson-Ternryd 1960) in 
which the basic standard errors of unit weight of common first order stereo 
scopic projection instruments have been determined from a great number of 
instruments. For the well known Wild autographs A7, for example, the stand 
ard error of unit weight of projector coordinate measurements was found to 
be 5 microns for base zero. This figure was found from 300 redundant obser 
vations (degrees of freedom). This means that the standard error of unit weight 
from a 9 points test sample of an autograph A7 (m = 4, n = 300) should not 
exceed 7.7 microns. If the result from a sample exceeds this value a possible 
decision to reject the instrument would be right in all but 5 percent of the cases. 
A more detailed example of tests performed on a first order stereoscopic 
projection instrument will be shown in Appendix I. 
1.22 Corrections (adjustment parameters) and Other Linear Functions of the Basic 
Observations 
From the adjustment procedure the assumed regular errors of the instrument 
(the parameters of the adjustment) will be found as linear functions of the 
basic observations (discrepancies). It is of great interest to find general rules 
for deciding when such adjustment quantities are large enough to require 
physical readjustment of the instrument. This is first of all a question of 
deciding the significance of deviations discovered in the instrument. All linear 
functions of the basic observations, for example radial distortion, can be 
considered similarly. 
For problems like these of determining significance, the 7-test of statistics 
is most common and can be used on condition that the error distribution of 
the basic observations is reasonably normal, see Appendix II and Diagram 1. 
For the practical application of the 7-test under these circumstances the 
standard error of the parameter (or any linear function) involved has to be 
computed first. This is done in the usual way by multiplying the basic standard 
error of unit weight by the square root of the weight number of the parameter 
or linear function.