i7000 and 110,000. The negative material was Gevaert Aviphot
Pan 30° film. For the further researches diapositives were used,
made on Gevaert plates.
In this report special attention will be given to the statistical
treatment of this experiment; as no use is made of the normal
distribution, which is almost exclusively used in surveying, it is
thought that this may be of interest. The judgment of the signals
was done in a stereoscopic model formed with diapositives. The
observer had to classify each signal in one of the following
groups, according to the brightness and dimension of the image
Brightness
Dimension
a.
good
good
b.
good
too large
c.
medium
good
d.
medium
too large
e.
poor
good
poor
too large
g-
invisible
This method will of course give rather subjective results, the
judging probably being dependent on the observer, contrast between
the signal and the surrounding part of the photograph a.s.o. Also
the opinion on the brigtness may be influenced by the dimension.
To abolish these objections to a certain extent the classification
was performed three times by different observers. Furthermore the
statistical treatment is done in a way as to nullify as far as possible
the influence of disturbing factors.
Tables 2a and 2b give the results of the classification. They have
been formed by first counting how many signals of a certain type
fell into the different classes for each observer; the entries in the
tables have been obtained by adding up the numbers of the different
observers, separately for the barren arable land (table 2a) and the
meadow-land (table 2b).
First an examination of the different colours will be made. This
can be tackled with the problem of m rankings [1]. The test will
be done by means of the number of signals, falling in class g
(invisible). Although even classification according to this charac
teristic is not free of personal influences, it seems the one least open
to them, the reason why this class is chosen. The test is performed
in the table 3a and 3b for respectively test-area a and b.
The left hand part contains, for the scale and dimensions specified
in the first columns per row the number of invisible signals, taken
from table 2, for the five different colours. In the right hand part
these numbers have been replaced by ranks, the smallest number
obtaining the rank 1 and so on. In case of equal numbers, so-called
ties, the ranks are equally divided. From this considerations it will
be evident that it is indifferent whether the sums (the entries in
table 2) are used or, for instance, the means of the observations;
iói