Geography Reference
In-Depth Information
Analysing the Accuracy of Answers
Monitored accuracy of answers was analysed separately for groups of stimuli with
the same position of legend, colour distance steps between classes and order of class
within the colour scheme, which corresponded to the correct answer to the
experimental task.
The less problematic composition of the map sheet was observed for the group of
stimuli, where the legend was placed on the right to the map sheet (no. IV).
However the observed 79 % accuracy does not exceed others groups of different
compositions obviously. Groups of stimuli with composition II, III, V and VI
showed the accuracy of 70-71 %. The lowest ratio of correct answers was observed
for the composition, where the legend is above the map sheet (I). To compare
accuracy of all groups see the Fig.
2a
.
Analysing the accuracy according to the colour distance steps between classes of
colour scheme brought prima facie evidence that the increasing colour distance has
a positive impact on the choropleth map readability (see Fig.
2b
). The lowest
overall accuracy (54 %) was observed for maps with the smallest colour distance
between areas (
E
00
¼
2). The best accuracy (82 %) was recorded for groups of
Δ
stimuli with
10.
The last experimental task was to find out how accurate are map-users while
matching different shades with the corresponding class in the legend. Particularly
problematic was to match the correct answer with classes in the
E
00
¼
Δ
of the
colour scheme. Less trouble causing was matching the lightest shades—the
observed accuracy for class A (corresponding to the very first class of the legend
and thus the most lightest) was higher than 90 % (see Fig.
2c
). The second class
(B) was also quiet easy to match (80 % accuracy). The most difficult was to match
correctly classes C and D (58 %, resp. 60 % accuracy). The 71 % accuracy for the
class F was rather surprising. It was assumed, that matching the extreme classes will
be easier, because by fast visual search one can quickly decide whether they can see
darker (or lighter) area than the one intended. However matching the lightest class
was much easier than to match the darkest class.
By deeper investigation of observed phenomenon it was found that low accuracy
observed for class F was caused by colour schemes with the small colour distance
between classes (
middle
'
'
2 and 4). Other colour schemes confirmed given assump-
tion—the extreme shades (classes A and F) are the most easier to match with the
legend (see Fig.
2d
).
Based on results above, it is apparently inappropriate to apply colour schemes
with the equal colour distance step between classes. Equal colour distance steps,
regardless their
Δ
E
00
value, are not ideal to allow easy discrimination between
'
Δ
E
00
¼
classes of the colour scheme.
In order to find out better solution three more colour schemes were designed and
tested. Optimized schemes are characterized by unequal colour distance steps with
higher
inner
'
E
00
values differentiating middle classes. For evaluation of optimized
Δ
