|
Table 5.5 Preliminary Data Analysis- Site Design (DES)
Step One - Exploratory
Indicators Wording
EFA
Loadings
DES1
This site is visually attractive
.78
DES2
This site is easy to use
.85
DES3
This site is easy to navigate
.86
DES4
This site quickly loads all the text and graphics
.84
DES5
This site provides me with sufficient information
.84
DES6
This site provides me with accurate information
.85
DES7
This site provides me with up-to-date information
.83
DES8
This site is quick and easy to complete a transaction
.80
Reliability
0.94
KMO
.901
Variance Explained
69.08
Bartlett’s
.000
Step Two – Confirmatory
Indicators CFA
Stand. Loadings
Stand.
Error
Critical
Ratio
AVE CR
DES1 .81
--
--
.64
.914
DES2 n/a
n/a
n/a
DES3 .77
.049
25.21
*
DES4 .80
.060
21.95
*
DES5 .93
.066
22.95
*
DES6 .82
.059
22.57
*
DES7 n/a
n/a
n/a
DES8 .77
.061
20.90
*
Model Fit
Statistic
Model Fit
Statistic
Chi-square 5.527
TLI
.999
df
4
CFI
.998
P .237
RMSEA
.020
GFI .998
RMR
.015
*
p
< .001
5.5.5 Preliminary Analysis – E- Service Quality
The e-service quality construct was measured using twelve items. These were subjected to the
two-step preliminary data analysis process shown in Figure 5.1, the results of which are
displayed in Table 5.6. Evaluation of the correlation matrix through the KMO and Bartlett’s
DES1 DES2 DES3 DES4 DES5 DES6 DES7 DES8
DES1
1.00
DES2
.767
1.00
DES3
.680
.813
1.00
DES4
.551 .662 .739 1.00
DES5
.547 .593 .645 .750 1.00
DES6
.571 .607 .625 .662 .767 1.00
DES7
.566 .601 .627 .595 .664 .735 1.00
DES8
.544 .587 .594 .590 .603 .641 .760 1.00
223
Test results in a high KMO statistic (.918) and a significant probability level (p< .001) for the
Bartlett’s test. These results indicate that sufficient correlations were found within the
correlation matrix for factor analysis to proceed. In addition, bivariate correlations were
inspected and all coefficients fell within the acceptable range for factor analysis of .30
and .90. EFA was then conducted which indicated that REL1 had a cross-loading problem.
After eliminating this item, EFA produced a three-factor structure for the eleven remaining
items used to measure e-service quality. The factor loadings ranged from .70 to .88 and the
cumulative variance explained by three factors was 79.59%, exceeding the recommended
criterion of 60% (Hair
et al.
2006). Cronbach’s alpha of .93 was then computed indicating
good reliability of the scale.
The first factor, composed of four items, accounted for 28.30% of the variance explained.
The four items loading on this factor reflected reliability with item loadings ranging from .85
to .92. Cronbach’s alpha for Factor 1 was .92, indicating good reliability of the scale.
Accounting for 27.78% of the variance explained, Factor 2 included four items related to
responsiveness. Factor loadings of these four items ranged from .85 to .90 and Cronbach’s
alpha of .89 indicated an internal reliability. The three items loading on Factor 3 were related
to personalization. The variance explained by this factor was 23.51% with strong item
loading ranging from .89 to .93, and Cronbach’s alpha was .90, exceeding the threshold
of .70 (Nunnally 1978). At this point, as all the remaining eleven items met the criteria of
Step One, they were retained for CFA analysis in Step Two.
The eleven items (retained from Step One) measuring e-service quality were subjected to
CFA in Step Two (refer Figure 5.1) of the preliminary analysis. The model fit indices
indicated that this measurement model did not fit quite well to the data (
χ
2
/df = 9.763 > 5,
224
RMSEA = .095 > .08, and RMR =.101> .05). In order to improve the model, indicators
which were related to larger reductions of chi-square or problematic standard residuals (2.5 as
cut-off) (Anderson and Gerbing 1988) were identified and eliminated one by one. After
removing item REL5 and item RESP4, an acceptable model was achieved. All overall
goodness-of-fit statistics were within acceptable ranges. For example,
χ
2
/df (2.183) was less
than 5, GFI (.989), TLI (.993) and CFI (.996) were above .90, RMSEA (.035) was less
than .08 and RMR (.044) was less than .05. Parameter estimates of the final measurement
model were inspected and no problems were found. Table 5.6 presents the second-order
measurement model in which
e-service quality was the second-order construct with three
first-order constructs (reliability, responsiveness, personalization) as indicators. The
standardized factor loadings of the three first-order constructs used to measure the second-
order construct (e-service quality) ranged from .75 to .83, all exceeding the preferable
criterion of .70 (Hair et al., 1995). The AVE for the second-order construct was .64,
exceeding the recommended level of .50 (Hair
et al.
2006). The CR of the second-order
construct was .840, exceeding the threshold of .70 (Nunnally 1978). More specifically, by
examining each first-order construct and its corresponding indicators, it can be found that the
standardized factor loadings of all items were relatively high and significant, ranging
from .82 to .91, and the AVE for all first-order constructs exceeded the recommended
criterion of .50, ranging from .73 (responsiveness) to .80 (reliability). CR of all first-order
constructs exceeded the recommended benchmark of .70, ranging from .891 (responsiveness)
to .924 (reliability).Therefore, items REL2, REL3, REL4, RESP1, RESP2, RESP3, PERS1,
PERS2, and PERS3 were considered to constitute a reliable and valid first-order
measurement scale. Taken together, this three-factor model structure was retained for
constructing the overall measurement model.
225
Do'stlaringiz bilan baham: |
