Perform Factor Analysis (Efa) for Independent Variables

Use Cronbach's Alpha reliability coefficient to eliminate inappropriate variables. Scales are accepted when the Cronbach's alpha coefficient is from 0.6 to 0.7 in the case of completely new research or new in the research context; Cronbach's alpha from 0.7 to 0.8 is acceptable and the best is from 0.8 to 0.9 (Nunnally & Burnstein, 1994). In addition, if the total item correlation coefficient of an indicator is greater than 0.3, that indicator is retained. But on the contrary, if a variable has a total item correlation coefficient of less than 0.3, it will be considered a garbage variable and will be eliminated from the model due to its poor correlation with other variables in the model.

Table 2.5. Results of Cronbach's Alpha analysis

Group

Maybe you are interested!

• Exploratory Factor Analysis Efa For Independent Variables
• Exploratory Factor Analysis Efa Of Independent Variables
• Summary Table of Exploratory Factor Analysis Results Efa
• Exploratory Factor Analysis (EFA) Results for Factors Affecting Customer Satisfaction
• Checking the Reliability of the Factor Scale Before Conducting Exploratory Factor Analysis Efa

Subjective standards

(Source: 2020 survey results)

Customers' attitudes towards buying cosmetics online

- Usefulness

The first test results show that all observed variables have appropriate total variable correlation coefficients (≥ 0.3). Cronbach's Alpha coefficient = 0.776 > 0.6, so it meets the reliability requirements. In which, variable HI4 (I buy cosmetics online at Myfaha because the product is cheaper) has a total variable correlation coefficient of 0.066 < 0.3, so the author removed this variable from the research model.

Then, the author conducted the second Cronbach's Alpha test, the test results showed that the observed variables all had appropriate total variable correlation coefficients (≥ 0.3). Cronbach's Alpha coefficient = 0.877 > 0.6, so it met the reliability requirements, in which the total variable correlation coefficients of the observed variables were all greater than 0.3, so the author did not exclude any variables from the research model.

- Ease of use

The test results show that the two observed variables SD1 (I buy cosmetics online at Myfaha because it is easy to operate on the website) and SD2 (I buy cosmetics online at Myfaha because it is easy to operate when making transactions and payments) with Cronbach's Alpha value = 0.509 < 0.6 do not meet the reliability requirements. Therefore, the author removed these two variables from the research model.

- Risk awareness

The test results show that the observed variables all have a total variable correlation coefficient.

appropriate (≥ 0.3). Cronbach's Alpha coefficient = 0.777 > 0.6 so it meets the reliability requirements.

- Subjective standards

The test results show that the observed variables all have a total variable correlation coefficient.

appropriate (≥ 0.3). Cronbach's Alpha coefficient = 0.703> 0.6 so it meets the reliability requirements.

- Interactivity, advertising

The test results show that the observed variables all have a total variable correlation coefficient.

appropriate (≥ 0.3). Cronbach's Alpha coefficient = 0.604 > 0.6 so it meets the reliability requirements.

- Control purchasing behavior

The test results show that the observed variables all have a total variable correlation coefficient.

suitable (≥ 0.3). Cronbach's Alpha coefficient = 0.933 > 0.6 so it meets the reliability requirements.

- Faith

The test results show that the observed variables all have a total variable correlation coefficient.

appropriate (≥ 0.3). Cronbach's Alpha coefficient = 0.730 > 0.6 so it meets the reliability requirements.

Thus, after Cronbach's Alpha test, there is 1 observed variable belonging to the usefulness scale (HI4: I buy cosmetics online at Myfaha because the product has a cheap price).

more) and 2 variables related to ease of use (SD1: I buy cosmetics online at Myfaha because it is easy to operate on the website; SD2: I buy cosmetics online at Myfaha because it is easy to operate when making transactions and payments) belonging to the attitude scale need to be removed before being included in the EFA exploratory factor analysis. The statistical table of the final test results of each group of variables is as follows:

Table 2.6. Results of Cronbach's Alpha analysis

STT

(Source: 2020 survey results)

2.2.3. Exploratory factor analysis (EFA)

The conditions for exploratory factor analysis must satisfy the following requirements:

Firstly, the KMO value ≥ 0.5 and the significance level of Bartlett's test is based on the Sig. value ≤ 0.05.

Bartlett's test is a statistical measure used to test the hypothesis that variables are not correlated in the population. The necessary condition for applying factor analysis is Sig.

≤ 0.05 the variables must be correlated with each other (Hoang Trong & Chu Nguyen Mong Ngoc, 2005). Therefore, if the test shows no statistical significance, factor analysis should not be applied to the variables under consideration. As for KMO (Kaiser-Meyer-

Olkin) is an index used to examine the appropriateness of factor analysis. If the KMO value is large (0.5 ≤ KMO ≤ 1), factor analysis is appropriate, but if this value is less than 0.5, factor analysis may not be appropriate for the data.

Second, the Eigenvalue > 1

The Eigenvalue represents the amount of variation explained by each factor. Only factors with an Eigenvalue greater than 1 are retained in the analysis model. Factors with an Eigenvalue less than 1 will not summarize information better than an original variable, because after standardization each original variable has a variance of 1.

Third, total extracted variance >= 50%

Variance Explained Criteria is the percentage of total variance explained by each factor. If the variance is considered to be 100%, this value indicates how much factor analysis explains. The total variance extracted must be at least 50% for factor analysis to be considered appropriate (Anderson & Gerbing, 1988).

Fourth, factor loading coefficient > 0.5

Factor loading is an indicator that shows the close relationship between the observed variable and the factor. According to Hair & colleagues (1998), factor loading is an indicator to ensure the practical significance of EFA: Factor loading > 0.3 is considered to reach the minimum level; Factor loading > 0.4 is considered important; Factor loading > 0.5 is considered to have practical significance. In this study, the researcher chose Factor loading ≥ 0.5. If any observed variable has a factor loading less than 0.5, it will be eliminated.

2.2.3.1. Perform exploratory factor analysis (EFA) for independent variables

Table 2.7. KMO and Bartlett's test results of independent variables

KMO and Bartlett's Test

(Source: 2020 survey results)

The results of KMO and Bartlett's tests are shown in Table 2.7, showing that the conditions are satisfied. With a coefficient of 0.5 ≤ KMO = 0.597 ≤ 1, it shows that exploratory factor analysis is accepted with the research data set. And Bartlett's test has a value of Sig. = 0.000 < 0.05, meaning that the observed variables have a linear correlation with the representative factor.

Table 2.8. Results of exploratory factor analysis EFA of independent variables

Variable

observe

(Source: 2020 survey results)

Eigenvalue of independent variable = 1.556 ≥ 1 means that 6 factors are extracted that best summarize the information. Total Variance

Explained) = 67.098% ≥ 50% shows that the variables mentioned in the reliability factor explain 67.098% of the variation in the data. The factor coefficients of the observed variables are all greater than 0.5 so they should be retained.

2.2.3.2. Conduct EFA analysis for dependent variable

Table 2.9. KMO and Bartlett's test results of dependent variable

KMO and Bartlett's Test

(Source: 2020 survey results)

The results of KMO and Bartlett's tests are shown in Table 2.9, showing that the conditions are satisfied. With a coefficient of 0.5 ≤ KMO = 0.677 ≤ 1, it shows that exploratory factor analysis is accepted with the research data set. And Bartlett's test has a value of Sig. =

0.000 < 0.05 means that the observed variables have a linear correlation with the representative factor.

area

Table 2.10. Results of exploratory factor analysis EFA of dependent variable

Dependent variable

(Source: 2020 survey results)

The Eigenvalue of the dependent variable = 2.023 ≥ 1 means that a factor has been extracted that best summarizes the information. The Total Variance Explained = 67.426% ≥ 50% shows that the variables mentioned in the reliability factor explain 67.426% ≥ of the variation in the data. The factor coefficients of the observed variables are all greater than 0.5 so they should be retained.

Based on the results of the EFA exploratory factor analysis of the independent and dependent variables, it shows that the EFA model is suitable for further research. Therefore, the author does not remove any observed variables from the factor model and does not continue to analyze the reliability of Cronbach's Alpha to test the reliability of the variables.

As a result of the final rotation matrix, we have the following redefined factors:

Table 2.11. Defined factors

STT

(Source: 2020 survey results)

2.2.4. Correlation analysis (Person)

The Pearson correlation coefficients of the independent variables usefulness (HI), perceived behavioral control (KS), interactivity/advertising (QC), trust (NT), perceived risk (RR) with the dependent variable intention (YD) are less than 0.05. Thus, there is a linear relationship between these independent variables and the dependent variable intention (YD). In which, the relationship between the independent variable usefulness (HI) and the dependent variable intention (YD) has the strongest correlation with the coefficient r of 0.508. And the relationship between the independent variable perceived risk (RR) and the dependent variable intention (YD) has the weakest correlation with the coefficient r (- 0.189).

The Pearson correlation coefficient between the dependent variable intention (YD) and the independent variable subjective norm (CM) is greater than 0.05, so there is no linear correlation between