36
(a.) Administration of alcohol after three series of measurement. The doses of 2 ml cognac per kg body
weight were given and blood sampling taken between the series.
(b.) Administration of alcohol at dinner the evening preceding the measurements. Doses were given
according to individual taste and blood samplings were taken during dinner and during the first hours
of work the following day.
Fifteen values of the standard deviation for each operator were obtained each day. The differences
between these values were tested according to Bartlett’s test. See tables 4.11:1 and 2 in the Appendix.
The precision shows a great variation in investigation (a) while it is more uniform in investigation (b).
The lack of uniformity in the first case may depend upon the degree of training. See fig. 4.11:1 in the
Appendix. Most of the lack of uniformity must be caused, however, by factors other than experience and
alcohol intake. In investigation (a) the correlation coefficients between blood alcohol concentration and
standard deviation for each day and each operator were computed. See table 4.11:3. In investigation
(b) a pooled standard deviation for each day and each operator, one value while influenced by alcohol
and the other without this influence. The differences between these values were tested with an F-test.
See table 4.11:4. In figures 4.11:2 through 9 in the Appendix results of the tests are shown graphically.
In conclusion, the standard deviation of one setting in x, y and z in the model was approximately
2 — 3 am on the image scale. The standard deviation of one parallax correction in a Zeiss stereocompa
rator 18 X 18 cm was approximately 3—4 um on the image scale. Two of the operators tested were less
experienced than the others.
The variation in the precision during a day is clear. In this investigation it appeared that alcohol in
take had no systematic effect on the variation.
4.12 The Effect of Smoking
As in the previous investigations, the precision was determined as a standard deviation of one setting
in a series of 25 replicated settings. There were two types of smoking with three operators taking part in
the tests.
(a.) Eight series with smoking between every second series.
(b.) Eight series with smoking between series nr 6 and 7.
The variation of the precision has been studied with Bartlett’s and F-tests. The variation of the mean
has been studied with variance analysis and t-tests.
The results of these tests show that there is a lack of uniformity in standard deviations and indicates
no significant connection with smoking.
Reference, 4.1:1. Torlegard, K.: Investigations of Setting Precision under Various Conditions in Photo-
grammetry, Svensk Lantmateritidskrift. Congress Number 1964.
4.2 Tests of Coordinatographs
The geometrical quality of coordinatographs was tested with methods similar to those used for tests
of comparators. See reference 4.2:1. With the coordinatograph a regular grid net of 25 points was plotted
on a sheet of material free from shrinkage. This sheet was then rotated through a right angle and the
coordinates of the plotted points were measured with the coordinatograph. The sheet was rotated
through another right angle, measured again, then rotated again and measured. Using the points
originally plotted as given data, an adjustment of the discrepancies in the following positions, according
to the method of least squares, revealed the most important regular errors of the coordinatograph, i.e.,
affine deformations and lack of orthogonality. The formal standard error of unit weight is more or less
a precision determination as no absolute data were given. Standard deviation of unit weight might be a
better expression here. For the practical application of the coordinatograph to photogrammetric plotting
this is less important because possible absolute scale errors are compensated by the scale of the model.