SOLVING GEOSPATIAL TASKS IN 2D AND 3D VISUALIZATIONS OF MAPS - Information Visualisation: Techniques, Usability and Evaluation - page 236

Accuracy
The distributions of observed accuracies and times were tested for
normality using a Shapiro-Wilk test with a significance level of a Į=0.05.
It revealed that accuracy was not normally distributed in the 45 ͼ condition
(n=36, W=0.91, p=0.005), but was in the 55 ͼ condition (n=36, W=0.95,
p=0.131) and 65 ͼ condition (n=36, W=0.95, p=0.074). Fig. 12.5(a) shows
box-plots of accuracy per block and per condition. It suggests that median
accuracy is at comparable levels for the 45 ͼ and 55 ͼ conditions, with a
salient drop in the 65 ͼ condition.
To test the hypothesis of equal means of the accuracies for the 45 ͼ , 55 ͼ
and 65 ͼ conditions, a Kruskal-Wallis test was employed. The statistics
F 2 =11.09, DF=2, p=0.0039 allows rejection of this hypothesis with an
error of Į=0.05.
In a post-hoc test, the 65 ͼ condition was contrasted against the
remaining two conditions employing a Bonferroni method to maintain an
overall confidence of 0.95. It showed that users' accuracy was
significantly reduced in the 65 ͼ condition, compared with the 55 ͼ
condition (p=0.0039) and 45 ͼ condition (p=0.0117).
Response times
The box-plot in Fig. 12.5(b) suggests that there were no differences in
response times for increasing slant angles. For all slant angles, observed
times were found not to be normally distributed.
The results of the Shapiro-Wilk tests were n=36, W=0.85, p=0.0002 in
the 45 ͼ condition; n=36, W=0.82, p<0.0001 in the 55 ͼ condition; and
n=36, W=0.82, p<0.0001 in the 65 ͼ condition. The results of a Kruskal-
Wallis test for comparison of the mean times in different slant angles did
not reveal any significant differences (F 2 =0.54, DF=2, p=0.7648).
Referring to the mean times in blocks, Fig. 12.5(b) suggests no overall
learning.