Understanding consumer online shopping behaviour from the perspective of transaction costs

Table 5.20 Results of Harman’s Single Factor Test 
 
Model 
χ
² (
p
-value) df 
GFI CFI 
TLI RMSEA RMR 
Δχ
² 
Δ
df
Original 
twelve-factor 
model 
3839.698 
(<0.001) 
1779 .932 .963 
.959 .035 
.102
Single-factor 
model 
5378.572 
(<0.001) 
2121 .795 .823 
.835 .289 
0.386 1538.874
*
342 
 
5.7 STRUCTURAL MODEL 
After testing the new overall measurement model by conducting a CFA, an adequate fit and 
satisfactory construct validity were achieved, which enabled an examination of relationship 
among sixteen important constructs. In this section, the structural model of this study was 
tested using SEM (Anderson and Gerbing 1988). SEM was particularly appropriate for the 
study of multiple dependence relationships such as those investigated in this study. 
The software package utilized for SEM in this study is the AMOS 21.0 statistical package 
due to its user friendliness. AMOS can link data directly to SPSS and provides a graphical 
user interface that allows configuring path diagrams, calculating model fit, and estimating 
parameters. For these reasons, and due to its increasing acceptance within the marketing 
domain, AMOS 21.0 was the chosen analytical tool to obtain the results which address the 
hypotheses of this study. As presented in Chapter 4, the five basic-building blocks in SEM 
suggested by Schumacker and Lomax (2004) were used to test the proposed models in this 

---

257
study, which included: specifying the model, assessing the identification of the model 
equations, estimating the proposed model, evaluating the results for goodness-of-fit, and 
interpreting and modifying the model.
The structural model is represented by Figure 5.2, which included the twelve major 
constructs. Compared to the research model shown in Figure 3.1, the structural model 
excludes the two antecedent variables (i.e., site design and environmental uncertainty) and 
the two moderators (i.e., consumer’s risk-bearing propensity and perceive enjoyment of 
online shopping). The structural model is used to test the causal research hypotheses (H1a, 
H1b, H1c, H2a, H2c, H2d, H3a, H3b, H4a, H4b, H4c and H5) of this study (H2b and H3c 
were not tested due to unsatisfactory discriminant validity of site design and environmental 
uncertainty). Tests of the partial mediation, moderators and difference across product 
categories will be discussed in the next sections. 
 

---

258
H4a: 
−
H4c: 
−
H4b: 
−