Information Technology Reference
In-Depth Information
accounts for roughly half of their attributes (17 of 37). The remaining five partic-
ipants (1,2,4,6,10) have two different, complementary models, i.e., the number of
attributes in the first model (26) is comparable to the number of attributes in the
second model (19). This shows that diversity is prevalent. Half of the participants
even hold two different views, explaining subgroups of attributes.
Ta b l e 3 . 4
Number of attributes explained by the two views for the ten participants of the
study.
Subj.
Total View a View b Remain
1
13
5
4
4
2
13
6
4
3
3
14
5
-
9
4
10
5
4
1
5
12
8
-
4
6
11
6
4
1
7
13
-
-
13
8
11
-
-
11
9
11
4
-
7
10
10
4
3
3
All together, 13 different views emerged from the ten individuals. These views
may partly overlap, which motivated us to group similar views and identify the
major diverse of this user group.
3.3.4
Assessing the Similarity between Different Views
In grouping the diverse views one has to derive a distance measure that reflects the
degree of dissimilarity between configurations. Each configuration can be regarded
as a
NxK
matrix, where
N
is the number of stimuli and
K
the number of dimensions
of the configuration space. The distance between configurations
X
n
and
X
m
can be
calculated using the
MATFIT
procedure, developed by Ramsay (1990).
MATFIT
seeks for a transformation matrix
M
that minimizes the distance measure:
d
2
t
=
trace
[(
X
n
M
−
X
m
)
](
X
n
M
−
X
m
)]
(3.3)
An arbitrary
KxK
transformation matrix M was applied. The procedure was repeated
with the matrices in reverse order as a means to calculating both distances: with
X
n
as independent and
X
m
as dependent, and vice versa. The resulting distances were
visualized in three dimensions using the program
XGms
(Martens, 2003). A hier-
archical (minimum variance) clustering algorithm was applied to the 3-D configu-
ration (a cluster is denoted by the lines connecting the different views). Figure 3.3
represents a 2-D perspective on the 3-D configuration of individual models. Note
that the distances in this 2D perspective do not necessarily reflect the true distances
Search WWH ::
Custom Search