CHAPTER 4 DATA ANALYSIS RESULTS AND DISCUSSION
4.2 Data Analysis
4.2.4 Validity Test
4.2.4.3 Confirmatory factor analysis (CFA)
Confirmatory factor analysis (CFA) is an important step before conducting an integrative structural model analysis. It is mainly used to test the relationship between observed data and latent variables. Based on the factor groupings
obtained from the exploratory analysis, the researcher needs to further examine the
"quality" of the observed data, i.e., validity testing. In the following, the structural validity, convergent validity, and discriminant validity of the model constructed in this study are examined.
1) Structural validity
In this paper, the validity of the constructed measurement model is tested using Mplus8.3. It was determined whether the observed data had significant loadings on the factors they were used for, and whether there were factors that were not relevant and did not have significant loadings. In general, if significance is shown and the standardized loading coefficient value is > 0.70, a strong correlation is indicated. Conversely, if it is not significant or the value of the standard loadings coefficient is < 0.4, the correlation between the analyzed term and the factor is weak and the removal of the observation should be considered. the relationships between the observations and the variables obtained in this study are shown in Figure 4.1.
In general, if it shows significance and the standard loading coefficient value is between 0.5 and 0.95 (Ning, 2017), it indicates a strong correlation. Conversely, if it is not significant or the standard loadings coefficient value is < 0.45 (Hou, J.T., 2004), it indicates a weak correlation between the analyzed term and the factor and the removal of the observation should be considered. the relationships between the observations and the variables obtained in this study are shown in Figure 4.1.
It is clear from Figure 4.1 that the values of the factor loading coefficients between each question item and the measured factor are in the range of 0.686-0.911, all of which meet the requirements, and all of the measured items show a significance at the 0.001 level (P < 0.001), so there is a good correspondence between the observed quantity and the factor and a good convergent validity.
Figure 4.1 Structural Validity Analysis of the Observed Model
2) Convergent validity
Convergent Validity is the validity of convergence. Convergent validity refers to the similarity of the results when different measures are used to measure the same feature. That is, the same characteristic properties should converge in different measurement methods. This study examined convergent validity by constructing reliability (CR) and average variance extracted values (AVE). Based on
the recommendation of T.S. Rong (2009), CR > 0.7 and AVE > 0.5 were considered as the criteria to be met. The scale and the factor loadings, combined reliability (CR) and average variance extracted (AVE) values for each dimensional observation in this study are shown in Table 4.6.
Table 4.6 Results of Convergent Validity Test Latent
Variables
Observed Variable
Factor Loading Coefficient
S.E. P CR AVE
EOU EOU1 0.835 0.015 *** 0.86 0.62
EOU2 0.853 0.014 ***
EOU3 0.727 0.021 ***
EOU4 0.717 0.022 ***
USE USE1 0.741 0.022 *** 0.832 0.55
USE2 0.727 0.022 ***
USE3 0.751 0.021 ***
USE4 0.756 0.021 ***
INT INT1 0.792 0.017 *** 0.88 0.6
INT2 0.832 0.015 ***
INT3 0.766 0.019 ***
INT4 0.778 0.018 ***
INT5 0.686 0.023 ***
TS TS1 0.745 0.02 *** 0.874 0.58
TS2 0.732 0.021 ***
TS3 0.803 0.017 ***
TS4 0.778 0.018 ***
TS5 0.761 0.019 ***
PQ PQ1 0.819 0.016 *** 0.883 0.65
PQ2 0.827 0.016 ***
PQ3 0.796 0.017 ***
PQ4 0.792 0.018 ***
Latent Variables
Observed Variable
Factor Loading Coefficient
S.E. P CR AVE
PV PV1 0.9 0.009 *** 0.943 0.81
PV2 0.881 0.01 ***
PV3 0.911 0.008 ***
PV4 0.9 0.009 ***
KS KS1 0.743 0.018 *** 0.95 0.66
KS2 0.76 0.017 ***
KS3 0.791 0.015 ***
KS4 0.786 0.016 ***
KS5 0.828 0.013 ***
KS6 0.852 0.012 ***
KS7 0.85 0.012 ***
KS8 0.813 0.014 ***
KS9 0.817 0.014 ***
KS10 0.846 0.012 ***
SS SS1 0.857 0.012 *** 0.94 0.76
SS2 0.89 0.009 ***
SS3 0.903 0.009 ***
SS4 0.86 0.011 ***
SS5 0.849 0.012 ***
Note: ***p < 0.001
From the results of the above table, it can be seen that the values of the factor loadings of the question items of each dimension of the scale are between 0.686-0.911, which indicates high convergent validity; the values of CR of each dimension are taken between 0.832-0.95, which are greater than 0.7 to reach the standard; the values of AVE are taken between 0.55-0.81, which are greater than 0.5 to reach the standard, and the probability of significance p < 0.001. This indicates that there is a significant relationship between the latent variables of the scale and the
variables of each measurement index, and the convergence of the scale structure model in this paper is very good.
3) Discriminant validity
Discriminant validity refers to the fact that the observed data should be distinguishable when different constructs are measured using various methods, thus reflecting the difference between the structure and other structures. Discriminant validity is usually tested using the square root of the AVE versus the correlation value. When the square root of the AVE is larger than the value of the correlation coefficient, then the discriminant validity is good.
The results of the validity tests for differentiation of the data on this study scale are shown in Table 4.7. The main body of the table shows the correlation coefficients between each latent variable, the diagonal line is the AVE value, and the last row is the square root of the AVE taken.
Table 4.7 Results of the Discriminant Validity Test
Notes: **p < 0.01; the diagonal line is the amount of variance extracted from the AVE evaluation variance
From the above table, it is clear that there is a significant correlation between all variables (p < 0.01). And the absolute value of the correlation coefficient is less than the square root of the corresponding AVE, which means that there is a
M SD EOU USE INT TS PQ PV KS SS
EOU 3.655 0.815 0.62 - - - - - - -
USE 3.203 0.757 .487** 0.55 - - - - - -
INT 3.276 0.755 .486** .556** 0.6 - - - - -
TS 3.45 0.73 .495** .464** .543** 0.58 - - - -
PQ 3.366 0.809 .384** .414** .458** .410** 0.65 - - -
PV 3.485 0.967 .523** .473** .501** .485** .437** 0.81 - - KS 3.403 0.858 .472** .435** .553** .595** .416** .479** 0.66 - SS 3.749 0.829 .617** .504** .561** .606** .483** .561** .637** 0.76
Square root of AVE 0.79 0.74 0.77 0.76 0.81 0.9 0.81 0.87
certain correlation between each latent variable and a certain degree of differentiation between them, which means that the discriminant validity of the scale data is ideal.