Page 561 - Ebook HTKH 2024

GP1c 0.781
GP1d 0.672
GP1a 0.668
GP2c 0.774
GP2b 0.744
GP2a 0.720
GP2d 0.647
GP4b 0.899
GP4a 0.788
GP4c 0.669
GP4d 0.382
DB3 0.837
DB2 0.808
DB1 0.682

Eigenvalue 5.209 2.616 2.196 1.548 1.209
% of Variance 25.151 11.377 9.284 6.059 4.178
Eigenvalue is a commonly used criterion to determine the number of factors in
EFA analysis. There are five factors extracted based on the criterion of eigenvalue

greater than 1, so these factors summarize the information of 19 observed variables
included in EFA in the best way. The total variance extracted by these five factors is
56.048% > 50%, thus, the 5 extracted factors explain 56.048% of the data variation of
19 observed variables participating in EFA.
Factor Loading, also known as the factor weight, represents the correlation
relationship between the observed variable and the factor. According to Hair et al.
(2010) Multivariate Data Analysis, variables with loading coefficients from 0.5 are
observed variables with good quality, the minimum should be 0.3. The results of the

rotation matrix show that 19 observed variables are classified into 5 factors, all observed
variables have Factor Loading coefficients greater than 0.5 and there are no bad
variables.
5.4. CFA – Confirmatory factor analysis
Confirmatory Factor Analysis is one of the statistical techniques of structural
equation modeling (SEM). CFA allows us to test how well the observed variables

represent the factors (constructs). CFA is the next step of EFA because CFA is used to
confirm univariate, multivariable, convergent and discriminant validity of the scale.
Table 5. Confirmatory factor analysis
Chi-square = 270.67; df = 142; P = 0.000

CMIN/df = 1.906;
GFI = 0.924; TLI = 0.938; CFI = 0.949
RMSEA = 0.052

To measure the fit of the model to market information, we often use Chi-square
(CMIN), Chi-square adjusted for degrees of freedom (CMIN/df), Comparative Fit Index
(CIF), Tucker & Lewis index (TLI), Root Mean Square Error Approximation index
(RMSEA),… The results of CFA analysis show that the model has statistical

553