Information Technology Reference
In-Depth Information
5R su s
We first screened the data for its conformance with the assumptions of our sta-
tistical test. One assumption behind the use of ANCOVAs is that the variables
are normally distributed. Kolmogorov-Smirnov tests confirmed that the depen-
dent variables “percentage of correct answers” and “perceived diculty” met
this criterion (
p
=0
.
105 and
p
=0
.
722).
For each dependent variable, we ran a univariate ANCOVA with “relations
between elements” and “element separateness” as independent factors and “el-
ement interactivity” as a covariate. According to the respective hypothesis, the
percentage of correct answers and the perceived diculty were the dependent
variables. The ANCOVAs allow us to test the influence of the three predictor
variables (two independent variables and a covariate) on the dependent variables
as well as possible interaction effects. We use “element interactivity” as a co-
variate, because it is a continuous variable, which co-varies with the dependent
variables percentage of correct answers (
r
=
−
0
.
37,
p
=0
.
002) and perceived
di
culty (
r
=
0
.
61,
p<
0
.
001). Therefore, the covariate accounts for variance
in the dependent variables and the inclusion of the covariate can increase the
statistical power of the procedure.
−
5.1 Results for Hypothesis 1
ANCOVA results indicate that there is a effect of different types of relations
between model elements (order, concurrency, repetition and exclusiveness) on
cognitive diculty of understanding their relation (H1). While there is only a
trend for the percentage of correct answers (F
3,55
= 2.65,
p
=0
.
058), the effect
on perceived diculty is significant (F
3,55
= 4.20,
p
=0
.
010). According to our
results, Hypothesis 1 is supported concerning subjective (perceived) diculty,
but only tentatively concerning objective diculty (correct answers). Table 1
gives the percentage of correct answers and perceived diculty for each type of
relation. Order was the easiest relation (80% correct answers) with the lowest
subjective diculty (3.08), followed by concurrency (83%, 3.20). Exlusiveness
was most the most di
cult relation concerning correct answers (70%, 3.19) and
repetition was rated as the most di
cult by participants (71%, 3.58).
Table 1.
Results for Hypothesis 1 (influence of types of relations between model ele-
ments), scale for perceived diculty from 1=”very easy” to 7=”very dicult”)
type of relation percentage of correct answers perceived diculty
Mean
SD Mean
SD
order
85%
6.16
3.08
0.35
concurrency
83%
10.09
3.20
0.28
repetition
71%
11.40
3.58
0.33
exclusiveness
70%
23.73
3.19
0.43
