5.3 Factor analysis
5.3.2 Confirmatory factor analysis (CFA)
After conducting EFA to confirm the validity of the conceptual model with regard to the uncorrelated factors from the survey, the collected data reflected the theoretical model through 10 factors. However, although the factor score could possibly be determined simply by averaging the raw scores of the loadings of all items into the factors, this would not take into account each item’s weight load (Hair et al., 2010). Hence, it was necessary to conduct confirmatory factor analysis (CFA) to ensure that a specific set of factors could influence the response in a predictable manner.
Constructs No.
Items
Mean(SD) Skewness Kurtosis Cronbach’s Alpha
Government support 8 0(1) -1.593 6.662 0.908
Sufficient IT knowledge/skills
4 0(1) -0.339 0.335 0.788
Insufficient facilitating conditions
4 0(1) -1.293 4.014 0.787
Intention of e-commerce adoption
3 0(1) -0.433 1.490 0.857
Acceptance and
adoption of e-commerce
4 0(1) -0.108 -0.740 0.822
Convergent validity refers to the degree to which measurement items of the same construct show a converged relationship, with this indicated by a proportion of variance which is shared among the items. In the current study, the measurement model assessment using CFA was conducted to confirm the convergent validity of each latent variable and also to evaluate the latent structure to see whether it had substantial justification (Byrne, 2001). To ensure an adequate level of convergent validity, the standardized factor loading and average variance extracted (AVE) value should be over 0.50 (Hair et al., 2010; Fornell & Larcker, 1981). Furthermore, the value of Composite reliability (CR) was used to measures the reliability of the factors, and it should ideally be above the value of 0.70 (Hair et al., 2010). The current study also conducted an observation of discriminant validity by comparing the AVE value of each construct indicator with the variance shared between each indicator and other indicators of the model.
As shown in Table 5.6, the CFA model fit indices were calculated and presented by using a combination of various types of fit indices in different categories as proposed in chapter 3 and it indicates the acceptable fit as all values of the statistical tests met the recommended level of fit criterion. The table provided the result of CFA model fit indices as follows: CMIN (x2) = 681.939; df = 625; p-value = 0.057; CMIN/df = 1.091;
GFI = 0.928; AGFI = 0.992; CFI = 0.992; NFI = 0.914; RMSEA = 0.020; and RMR = 0.015. Consequently, the results for all the fit indices of the CFA model revealed a good
model fit, based on the suggested parameters and supported the AVE analysis in confirming convergent validity.
Table 5.6: CFA model fit indices
Fit indices Recommended value
Structural model value
Results
Chi-square/degrees of freedom
(CMIN/df) ≤ 2.00 1.367 Acceptable
p-value ≥ 0.05 0.063 Acceptable
Goodness-of-fit index (GFI) ≥ 0.90 0.905 Acceptable
Adjusted goodness-of-fit index
(AGFI) ≥ 0.90 0.907 Acceptable
Comparative fit index (CFI) ≥ 0.90 0.964 Acceptable
Normed fit index (NFI) ≥ 0.90 0.958 Acceptable
Root mean square error of
approximation (RMSEA) ≤ 0.05 0.030 Acceptable
Root mean square residual
(RMR) ≤ 0.05 0.019 Acceptable
Table 5.7 presents a summary of the CFA results. The factor loadings of all 41 items had values above the recommended value of 0.50 (Hair et al., 2010). In addition, the composite reliability (CR) values of the constructs ranged from 0.76 to 0.89, satisfying the standard of at least 0.7, whereas the average variance extracted (AVE) values for each construct also met the threshold of AVE ≥ 0.5, thus suggesting good convergent validity (Hair et al., 2010; Fornell & Larcker, 1981). It implies that more than 50 percent of the each variances observed in these 41 items were accounted for by their indicators.
Table 5.7: Factor loadings, t-values, R2, CR, and AVE the measurement model
Variables B SE t-value R2 CR
(≥ 0.70)
AVE (≥ 0.05)
PE1 0.633 --- --- 0.401 0.81 0.68
PE2 0.616 0.099 9.633** 0.379 PE3 0.632 0.111 9.152** 0.399 PE4 0.584 0.112 8.715** 0.340
EE2 0.651 --- --- 0.424 0.83 0.60
EE3 0.754 0.098 10.707** 0.569 EE4 0.602 0.079 9.816** 0.362
SI1 0.673 --- --- 0.453 0.78 0.53
SI2 0.645 0.112 10.046** 0.416 SI3 0.643 0.090 10.584** 0.413 SI4 0.649 0.102 10.202** 0.422
IC1 0.667 --- --- 0.445 0.77 0.53
IC2 0.827 0.104 11.112*** 0.684 IC3 0.618 0.087 10.713*** 0.381 IC4 0.514 0.112 7.244*** 0.245
PR1 0.813 --- --- 0.661 0.84 0.66
PR2 0.783 0.057 16.717** 0.613 PR3 0.804 0.061 16.809** 0.646
IT1 0.757 --- --- 0.573 0.86 0.61
IT2 0.786 0.075 13.999** 0.618 IT3 0.625 0.072 11.191** 0.390 IT4 0.528 0.074 8.797** 0.238
FC1 0.703 --- --- 0.495 0.87 0.62
FC2 0.768 0.093 13.658 0.591
FC3 0.737 0.085 12.831 0.543
FC4 0.557 0.075 9.995 0.311
GOV1 0.679 --- --- 0.461 0.76 0.55
Variables B SE t-value R2 CR (≥ 0.70)
AVE (≥ 0.05) GOV2 0.777 0.082 13.925*** 0.604
GOV3 0.770 0.076 14.075*** 0.592 GOV4 0.819 0.086 14.761*** 0.671 GOV5 0.783 0.081 14.000** 0.613 GOV6 0.641 0.076 11.785** 0.410 GOV7 0.651 0.085 11.921** 0.424 GOV8 0.683 0.080 12.436** 0.467
INT1 0.800 --- --- 0.639 0.82 0.61
INT2 0.885 0.062 18.159 0.783 INT3 0.768 0.059 16.226 0.589
AA1 0.660 --- --- 0.435 0.89 0.72
AA2 0.795 0.101 11.992 0.633
AA3 0.738 0.102 11.608 0.545
AA4 0.683 0.108 10.874 0.467
*p-value < 0.05; ** p-value < 0.01; *** p-value < 0.001 (t-value > 2.58) --- No report on SE and t-value due to fixed parameters ---
Note: AA = acceptance and adoption of e-commerce; EE = effort expectancy; FC = facilitating conditions; GOV = government support; IC = perceived implementation cost; INT = intention of e-commerce adoption; IT = information technology (IT) knowledge and skills; PE = performance expectancy; PR = perceived risk; and SI = social influence
Figure 5.1 presents the CFA model by using the 41 items proposed from the EFA model. The oval shaped items on the right represent the ten factors of the study consisting of performance expectancy, effort expectancy, social influence, perceived implementation cost, perceived risk, sufficient IT knowledge and skills, government support, facilitating conditions, behavioral intention and acceptance and adoption of e-commerce. Co-variances between each of these factors are also drawn and the values are reported on the right side of the diagram. Each factor is represented by a number of measured variables or indicators designated by a box. These measured variables were captured in the questionnaire used by this study. Factor loadings for measured variables are also reported (the line between the oval and box).
Figure 5.1: The CFA model
Furthermore, the discriminant validity was also conducted to assess the extent to which a concept and its indicators differ from another concept and its indicators (Bagozzi, Yi & Phillips, 1991). According to Fornell & Larcker, (1981) the inter- construct correlations should be lower than the square root of the average variance extracted of each construct. Therefore, to test for discriminant validity, the square roots of AVE values and their correlation with the other factors were compared. As shown in Table 5.8, the square roots of AVE values were greater than the correlation between the variables, revealing good discriminant validity (Fornell & Larcker, 1981).
Table 5.8: Correlation coefficient matrix and square root of AVEs
PE EE SI IC PR IT GOV FC BI AA
PE 0.819
EE 0.326 0.768
SI 0.359 0.340 0.721
IC 0.352 0.338 0.310 0.742
PR 0.347 0.357 0.332 0.398 0.806
IT 0.385 0.333 0.438 0.395 0.304 0.787
GOV 0.336 0.404 0.381 0.460 0.323 0.396 0.735
FC 0.309 0.408 0.365 0.360 0.491 0.393 0.309 0.794
BI 0.334 0.338 0.304 0.450 0.455 0.338 0.388 0.334 0.781
AA 0.497 0.405 0.537 0.442 0.487 0.419 0.451 0.423 0.440 0.837
Note: AA = acceptance and adoption of e-commerce; EE = effort expectancy; FC = facilitating conditions; GOV = government support; IC = perceived implementation cost; BI = behavioral intention; IT = information technology (IT) knowledge and skills; PE = performance expectancy; PR = perceived risk; and SI = social influence.