Information Technology Reference
In-Depth Information
Ta b l e 3 . 3
Goodness of fit statistics for the two diverse models of participant one. At-
tributes (2,4,6,8,12,13) were adequately predicted (
R
2
>
.5
&
R
k
>
6
) by model 1. Attributes
(1,5,7,9,10,11) displayed the least fit (
R
k
<
4
) and were used to derive a second model. At-
tributes (5,7,9,10) were adequately predicted by model 2.
Attribute
Avg. Model Model 1
Model 2
2
)
R
2
R
2
R
2
No
Variance
(σ
R
k
R
k
R
k
1
2.6
0.47
2.2
0.47 2.2 0.36 3.3
2
3.8
0.89
7.3
0.89 7.3
3
0.6
0.73
4.1
0.73 4.1 0.56 2.6
4
1.9
0.98
18.6
0.98 18.6
5
3.7
0.49
2.3
0.49 2.3 0.95 13.7
6
2.2
0.99
40.5
0.99 40.5
7
1.7
0.48
2.4
0.48 2.4 0.99 39.6
8
6.3
0.93
9
0.93
9
9
4.1
0.63
4.8
0.63 4.8 0.99 40.1
10
4.5
0.26
2.1
0.26 2.1 0.61 6.5
11
3.9
0.08
0.9
0.08 0.9
12
1.9
0.88
6.8
0.88 6.8 0.48 2.8
13
5.6
0.99
50.4
0.99 50.4 0.85 5.5
configuration space (which is represented in the model parameters) displays virtu-
ally no change. In other words, the attributes that were removed (according to the
arbitrary criteria) had no contribution to the configuration space. Thus, the infor-
mation contained in these attributes is not modeled when attempting an averaging
analysis and therefore it is lost.
Out
of
all
the
attributes
that
were
not
adequately
predicted,
attributes
(1,5,7,9,10,11; in italics) displayed the least fit by model 1, i.e.,
R
k
<
4
.Thesewere
used to derive a second model. Out of them, only attributes (5,7,9,10) turned out to
be adequately predicted by model 2, using the same goodness of fit criteria as used
in model 1.
Figure 3.2 illustrates the different insights that the two diverse views bring. One
can note that the two views highlight semantically different attributes. Each attribute
is visualized as an arrow, i.e. a dimension, on which the relative positions of the
websites can be compared. The length of each arrow indicates the strength of the
attribute, reflecting the variance in the predicted attribute ratings for the different
stimuli; on some attributes all websites might be rated as 4 or 5 on a 7-point scale,
while others might make strong differentiations between sites, i.e. across the whole
range of the scale.
The first view provides overall three different insights. First, that the universi-
ties of Frankfurt, Manheim and Mainz are perceived as putting
less emphasis on
achievement
, as compared to the remaining five universities. This may be induced
by the websites but may also reflect prior beliefs of the individual. Second, the web-
sites of the universities of Mnchen, Aachen, Karlsruhe and Heidelberg have a more
professional layout
as opposed to the remaining four which have a more
playful
one.
Search WWH ::
Custom Search