Understanding consumer online shopping behaviour from the perspective of transaction costs

Table 5.4 Preliminary Data Analysis- Product Quality Concern (PQC) 
Step One - Exploratory 
Wording EFA 
 
Loadings 
PQC1 PQC2 PQC3 
PQC4 PQC5 
Difficult to be assured that the product will 
perform as well as it is supposed to 
.86 PQC1 
1.00 
Difficult to be assured that the product is 
reliable 
.91 PQC2 
.782 
1.00 
Worried that the product received cannot 
represent accurately 
.93 
PQC3 
.725 .825 1.00
Worried that the performance of the 
product cannot live up to what is promised 
.92 
PQC4 
.703 .772 .852 1.00
Worried that the quality of the product 
.88 
PQC5 .680 
.710 
.756 
.822 
1.00 
Reliability 
0.94 
KMO 
.874 
Variance Explained 
81.07 
Bartlett’s 
.000 
Step Two – Confirmatory 
Indicators CFA 
 
Stand. Loadings 
Stand. 
Error 
Critical 
Ratio
AVE CR 
PQC1 .81 
-- 
-- 
.75 
.937 
PQC2 .80 
.033 
32.80
*
PQC3 .89 
.040 
31.15
*
PQC4 .96 
.045 
32.17
*
PQC5 .85 
.043 
29.62
*
Model Fit 
Statistic 
Model Fit 
Statistic 
χ
2
1.452 
TLI 
.996 
df
1 
CFI 
.996 
P .228 
RMSEA 
.022 
GFI .995 
RMR 
.010 
* 
p 
< .001
5.5.4 Preliminary Analysis – Site Design 
Eight items (DES1 to DES8) were used to measure the site design
 
construct. The two-step 
preliminary data analysis process shown in Figure 5.1 was employed as a screening 
mechanism for data measured by these items (refer Table 5.5 for results). Evaluation of the 
correlation matrix through the KMO and Bartlett’s Test results in a high KMO statistic (.901) 
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 produced 

---

221
a single factor structure with strong item loading ranging from .78 to .85 and the variance 
explained was 69.08%. Cronbach’s alpha of .94 indicated good reliability of the scale. At 
this point, as all the items met the criteria of Step One, they were retained for CFA analysis in 
Step Two. 
In Step Two (refer Figure 5.1), CFA was conducted on the eight items retained from Step 
One. The model fit indices indicated that this measurement model poorly fitted the data (
χ
2
/df 
= 46.23 > 5, GFI = .787 < .90, TLI = .799 < .90, CFI = .857 < .90, RMSEA = .217 < .08, and 
RMR =.143 > .05). After the initial model assessment, the model was modified according to 
the modification indices to improve model fit (Hair et al., 1995). Indicators which were 
related to larger reductions of chi-square were identified and eliminated one by one. After 
deleting item DES2 and DES7, an acceptable model was achieved. For example, 
χ
2
/df (1.382) 
was less than 5, GFI (.998), TLI (.999) and CFI (0.998) were above .90, RMSEA (.020) was 
less than .08 and RMR (.015) was less than .05. The results indicated that the modified 
measurement model can be assessed as being adequate, and suggested that the modified 
model fitted the data better than the initial model. As displayed in Table 5.5, the standardized 
factor loadings of the remaining six items ranged from .77 to .81. All exceeded the ideal 
criterion of .70 (Hair
 et al.
2006). The AVE for this construct was .64, exceeding the 
recommended level of .50 (Hair
 et al.
2006). The CR was .914, exceeding the recommended 
criterion of .70 (Nunnally 1978). As a result, items DES1, DES3, DES4, DES5, DES6, and 
DES8 were considered to provide a reliable and valid measurement scale for site design. 
Moreover, these six items were retained for the overall measurement model. 

---

222