Information Technology Reference
In-Depth Information
mean mouse clicks of 38.563 (SE 3.235) for the
normal
condition compared to 43.109
(SE 3.036) for the
physics
condition.
There was also a significant main effect on the graph coloring,
F
(1
,
15)=11
.
614
,
p
<
0
.
01. Participants made significantly fewer mouse clicks for the
plain
condition than
colored
, with mean mouse clicks of 39.125 (SE 2.685) for the
plain
condition com-
pared to 42.547 (SE 3.367) for the
colored
condition.
5.2
Graph Layout Analysis
Layouts were analysed for stress and edge crossings. The results presented here are
represented as a ratio of changeinvalues compared to the starting values of the ini-
tial graph layout, i.e.
Result
=(
EndValue
StartValue
)
/
StartValue
.Using a ratio of
change allows comparisons between graph layouts of different sizes, numbers of clus-
ters, and vastly different initial stress and edge crossing values. In all conditions, stress
was higher after participants interacted with the graph. Of the 128 trials conducted in
the experiment, only two resulted in lower stress values than the initial conditions. How-
ever, there was a significant main effect on interaction mode,
F
(1
,
15)
,
p
<
0
.
05. The
physics based interaction produced layouts with significantly less stress than the normal
interaction mode, with mean stress change of 22.884 (SE 2.772) for the
physics
con-
dition compared to 95.763 (SE 14.955) for the
normal
condition. Graph coloring did
not produce a significant effect on graph stress, and no significant interaction between
interaction mode and graph coloring was found. Edge crossingswerealsohigher after
participants interacted with the graph. Sixteen of the 128 trials showed a reduction in
edge crossings.
There was a significant main effect on interaction mode,
F
(1
,
15)
,
p
<
0
.
05. The
physics based interaction produced layouts with significantly fewer edge crossingsthan
the normal interaction mode, with mean changeinedge crossings of 0.578 (SE 0.087)
compared to 2.811 (SE 0.538) for the normal interaction mode. Graph coloring did not
significantly affect edge crossings, and no significant interaction between interaction
mode and graph coloring was found.
−
5.3
Questionnaire Results
The results of the questionnaire comparing the physics interaction to normal interaction
are summarised in Fig.4.Significant results (
p
<.
05) were found for questions 1,
3, and 4, with participants giving higher scores for the physics condition in all three
questions. Participants overwhelmingly preferred the physics interaction, with 87.5%
of participants choosing physics as the preferred mode. Participants also responded
favourably to the graph coloring.Results for Q5 were 71.69 (SD 18.095), and Q6 68.19
(SD 15.753).
6
Discussion
The results of the user study show that the GION interaction technique is better in the
untangling task than existing interactive methods. Specifically, hypotheses H1, H2, H5,
