Information Technology Reference
In-Depth Information
unconstrained and constrained models are 415.62
(df = 137) and 426.65 (df = 142), respectively. Their
difference is 11.03, with 5 degrees of freedom. The
significant difference (at the 10% level) indicates
that moderating effects do exist. The χ
2
difference
test is used again to test for the moderating ef-
fects of individual paths, as displayed in Figure
1. However, the χ
2
statistics for the unconstrained
and the partially constrained models are compared
herein.
Partially constrained
means that only
the target path coefficients are set to be equal for
cross-group datasets.
and H
9
are supported), while the influence of
relational selling behavior on relationship quality
is stronger for the female group than for the male
group (H
8
is supported). The influences of exper-
tise and service recovery on relationship quality
are similar for both the male and female groups
(H
7
and H
10
are not supported). The above results
supporting H
6
, H
8
, and H
9
show weak moderating
effects (statistically significant at the 0.1 level).
Finally, Table 7 summarizes the test results.
Hypotheses H
2
and H
7
are not supported, maybe
because of the virtual quality of Internet services.
Unlike in other industries, ISP customers face
difficulty distinguishing the expertise of differ-
ent service providers. Moreover, the unsupported
H
10
may unveil an interesting phenomenon in that
the influence of service recovery on relationship
quality might be non-gender-specific, indicating
it is equally important for both genders.
empirical results
Based on good model fitness as described previ-
ously, Table 5 lists the empirical testing results.
Based on the entire sample (total group), one path
is not supported (H
2
is not supported), while the
remaining paths are all significant at the 0.01 level
(H
1
, H
3
, H
4
, and H
5
are supported).
Figure 2 and Table 6 list the further test of the
moderating effect for individual paths. From Table
6, the test results indicate that the influences of
perceived network quality on relationship quality
and of relationship quality on loyalty are stronger
for the male group than for the female group (H
6
disCussion and managerial
impliCations
A key challenge for research on relationship
quality is identifying and understanding how
managerially controlled antecedent variables
Table 5. Path coefficients and t values for three models
Total
Male
Female
Hypothesis
Standardized
coefficient
Standardized
coefficient
Standardized
coefficient
t
value
t
value
t
value
H
1
0.79***
14.99
0.85***
13.94
0.66***
7.17
H
2
0.03
0.36
0.05
0.49
0.03
0.28
H
3
0.36***
3.13
0.26*
1.70
0.58***
3.11
H
4
0.30***
3.97
0.36***
3.95
0.01
0.03
H
5
0.30***
2.78
0.33**
2.30
0.33*
1.92
Notes:
*** significant at the 0.01 overall significance level.
** significant at the 0.05 overall significance level.
* significant at the 0.10 overall significance level.
Search WWH ::
Custom Search