Factors affecting decisions on choosing English centers of students in Ho Chi Minh City

Tho, 2013). Also according to Nunnally and Bernstein (1994), states that we need to consider the correlation coefficient between observed variables and the sum of the variables of the scale, if the correlation coefficient between which is 0.3, then the variable is accepted. take. All variables accepted for reliability will continue to be explored by the Exploratory Factor Analysis - EFA. Exploratory Factor Analysis EFA EFA - Exploratory Factor Analysis: is an analysis method of multivariate analysis group that has interdependence among variables, that is, it relies on the correlation between variables, absolutely no dependent variables and Independent variables. Use the discovery factor analysis (EFA) method to reduce the number of k observed variables to F observed variables (where F < k) to make the observed variables more meaningful. Besides, when conducting the analysis, we can also evaluate two important values ​​of the scale: discriminatory and convergent values ​​(Nguyen Dinh Tho, 2013). In this analysis method, the index used to consider the appropriateness of factor analysis is KMO (Kaiser - Meyer - Olkin) value. If the KMO value is less than 0.5, it is possible to analyze the factor that is not suitable for the data, KMO must have a value from 0.5 to 1 to be appropriate. Bartlett's test tests the hypothesis of a correlation between zero observed variables in the population. According to Hoang Ngoc and Chu Nguyen Mong Ngoc (2008), if this test is performed and results are statistically significant (sig. < 0.05), the observed variables have a correlation in the overall. Some authors Mayers, LS, Gamst, G. and Guarino AJ (2000), stated that the method of extracting Principal Components Analysis along with Varimax rotation is used a lot in factor analysis, so the team will also use this method. We will continue to eliminate variables with factor loading of less than 0.5. However, we also need to consider the content value of the variable, if the weight of the variable is not too small but its content value is important, it should not be removed (Nguyen Dinh Tho, 2013). In the case of variables with the difference between the maximum factor load factor and any factor load factor 50% (Nguyen Dinh Tho, 2013). Analysis of Pearson-r correlation coefficient A statistical coefficient called Pearson correlation coefficient was used to quantify the degree of rigidity of the linear relationship between the independent and dependent variables. A positive correlation coefficient indicates that the two variables have an absolute positive relationship. If there is a strong correlation between two independent variables, multi-collinearity must be considered when regression analysis. In Pearson correlation analysis, there is no distinction between the independent and dependent variables, all of which are considered equally. Multivariate regression analysis Regression analysis: multiple linear regression analysis is a method using statistical techniques to estimate the impact equation between independent and dependent variables. There are two issues to note when using multivariate regression methods. First, between the independent and dependent variables is a correlation. Secondly, the statistical parameters that need to be noted are: Adjusted coefficient (R2): Independent variable (including sample size and dependent variable) will explain the measurement of variance of the dependent variable. The accuracy of the model and the prediction of the independent variables will be higher if this coefficient is higher. Test the compatibility level between the model and the data set: the model is tested for the meaning of statistics through the statistic value F. All Beta coefficients in the model are zero (hypothesis Ho). If, after testing, the significance level is less than 0.05, we can reject the Ho hypothesis. Beta coefficient (Standardized Beta Coefficient): can compare the effect of independent variables on the dependent variable thanks to the standardized regression coefficient. Testing the significance level of Beta coefficient: check this significance level by using statistical values. The Beta will be statistically significant if the significance is less than 0.05. Summary of Chapter 3: This chapter presents qualitative research and quantitative research. The student's decision-making scale consists of 38 questions in 10 elements and a "Decision" (6 questions) is considered a result of student decision. Data processing software SPSS version 18.0 is used to describe data, evaluate the reliability, validity of the measurement scale as well as perform other inference statistics. The research results after conducting data analysis will be presented in detail in the next Chapter 4. RESEARCH RESULTS Description of the research sample The author has conducted official research during the period from February 4, 2020 to April 25, 2020. Appendix X documents the official survey. Following are the survey results: Proceed to post the survey with the expectation of receiving 400 responses. The author collected 420 answers. During the import and processing process there are only 403 valid tables and 17 answers. The main reason is due to answering parameters that are either regular or the same. The statistical table (Chart X1) shows the sample structure that was investigated. The structure table shows that the majority of survey participants are students studying at the Foreign Trade University, Ho Chi Minh City Campus (33.1%). Other universities and colleges include University of Economics, Banking Universities, Vietnam National University - Ho Chi Minh City, Ho Chi Minh City Open University, Ho Chi Minh City University of Architecture, Ho Chi Minh City University of Arts, Ho Chi Minh City University of Foreign Languages and Information Technology, Ho Chi Minh City University of Industry, Ho Chi Minh City University of Information Technology, Ho Chi Minh City University of Law, Ho Chi Minh City University of Medicine and Pharmacy, Ho Chi Minh City University of Pedagogy, Ton Duc Thang University, University of Communications and Transportation, College Of Foreign Economic Relation and Posts and Telecommunications Institute of Technology. In addition, more than 50% of the surveyed students are sophomore and third-year students. The remaining 50% are freshmen, final-year students (fourth year, fifth year, sixth year) and freshly graduated students (not more than 2 years). Specific statistics are shown in chart X2. There are 298 female students participating in the survey (nearly 74%). The remaining belongs to male students and those who do not want to give information in this field. Figure 41 Research Sample 1 Source: Compiled by the author Figure 42 Research Sample 2 Source: Compiled by the author Analyze the reliability of the scale by Cronbach’s Alpha coefficients Cronbach’s alpha coefficient is a statistical test used to check the coherence and correlation between observed variables. This is an essential analysis step to eliminate garbage variables before using EFA. This relates to two aspects which are the correlation between the variables themselves and the correlation of the scores of each variable with the total score. Only variables with variable correlation - total correction (Corrected Iterm - Total Correlation) greater than 0.3 and Cronbach's Alpha coefficient greater than 0.6 will be considered acceptable and appropriate for inclusion in the Further analysis steps (Nunnally, 1978; Peterson, 1994; Slater, 1995). After the investigation, the subject conducted the reliability test of the scale that decided to choose a language center using the Cronbach’s Alpha coefficient. The results are as follows: Analytical results of the scale "Geographical location" Table 41 Geographical Location Reliability Statistics Cronbach's Alpha N of Items .627 4 Source: Compiled by the author Table 42 Geographical Location Item-Total Statistics Scale Mean if Item Deleted Scale Variance if Item Deleted Corrected Item-Total Correlation Cronbach's Alpha if Item Deleted Loca1 11.2655 2.434 .544 .441 Loca2 11.0199 2.626 .550 .445 Loca3 11.4864 2.718 .618 .410 Loca4 11.0347 4.143 .007 .794 Source: Compiled by the author The component "Geographical location" consists of four observed variables (Loca1, Loca2, Loca3, Loca4). After checking Cronbach's Alpha, the results showed that there were variables Loca4 "Being near tourist centers where there are foreign visitors." with Corrected Item-Total Correlation of 0.007 less than 0.3, so it was disqualified. Then the "Geographic location" scale has only three variables. The result is different from the result of Doan Thi Hue's author with students at Nha Trang University when the author has eliminated the "Location" variable, but quite consistent with the results of the previous authors about decided to choose a university like Nguyen Phuong Toan (2011), Tran Van Quy and Cao Hao Thi (2009), the results "Characteristics of the school: The school has a suitable position" have a positive impact on students’ decision. Analytical results of the scale "Marketing" Table 43 Marketing Reliability Statistics Cronbach's Alpha N of Items .828 4 Source: Compiled by the author Table 44 Marketing Item-Total Statistics Scale Mean if Item Deleted Scale Variance if Item Deleted Corrected Item-Total Correlation Cronbach's Alpha if Item Deleted Mkt1 10.4715 5.568 .624 .797 Mkt2 10.3697 5.189 .664 .779 Mkt3 9.9677 5.519 .632 .794 Mkt4 9.9032 4.829 .705 .760 Source: Compiled by the author The "Marketing" component has 4 variables including Mkt1, Mk42, Mk43 and Mkt4. The analytical results show that the Coefficient Cronbach’s Alpha = 0.828 is greater than 0.6 so the "Marketing" scale is reliable and is a good scale. Analytical results of the scale "Training program" The analytical results show that the Coefficient Cronbach’s Alpha = 0.729 so the "Program" scale is reliable and if any observed variable is removed, the Alpha coefficient is less than 0.729. Therefore all five observed variables in this factor are accepted and continue to be used in EFA factor analysis. Table 45 Training program Reliability Statistics Cronbach's Alpha N of Items .729 5 Source: Compiled by the author Table 46 Training program Item-Total Statistics Scale Mean if Item Deleted Scale Variance if Item Deleted Corrected Item-Total Correlation Cronbach's Alpha if Item Deleted Pro1 15.4119 7.521 .441 .704 Pro2 15.6005 8.111 .410 .711 Pro3 15.2978 7.220 .545 .659 Pro4 15.4665 7.011 .637 .622 Pro5 15.4739 8.444 .426 .705 Source: Compiled by the author Analytical results of the scale "Training quality" Table 47 Training quality Reliability Statistics Cronbach's Alpha N of Items .624 3 Source: Compiled by the author Table 48 Training quality Item-Total Statistics Scale Mean if Item Deleted Scale Variance if Item Deleted Corrected Item-Total Correlation Cronbach's Alpha if Item Deleted Qua1 7.3350 1.860 .463 .480 Qua2 7.9901 2.254 .363 .616 Qua3 6.9677 1.837 .477 .460 Source: Compiled by the author The results show that the Coefficient Cronbach’s Alpha = 0.624 is greater than 0.6, then the "Training quality" scale is decent. Therefore all observed variables in this factor are accepted and continue to be used in EFA factor analysis. Analytical results of the scale "Teachers" The results show that the Coefficient Cronbach’s Alpha = 0.611 is greater than 0.6, then the "Teachers" scale is meaningful. Therefore all three observed variables in this factor are accepted and continue to be used in EFA factor analysis. Table 49 Teachers Reliability Statistics Cronbach's Alpha N of Items .611 3 Source: Compiled by the author Table 410 Teachers Item-Total Statistics Scale Mean if Item Deleted Scale Variance if Item Deleted Corrected Item-Total Correlation Cronbach's Alpha if Item Deleted Tea1 7.6253 1.429 .348 .608 Tea2 8.5881 1.243 .446 .474 Tea3 8.2233 1.134 .471 .433 Source: Compiled by the author Analytical results of the scale "Tuition" Table 411 Tuition Reliability Statistics Cronbach's Alpha N of Items .665 4 Source: Compiled by the author Table 412 Tuition Item-Total Statistics Scale Mean if Item Deleted Scale Variance if Item Deleted Corrected Item-Total Correlation Cronbach's Alpha if Item Deleted Fee1 10.3648 4.058 .347 .663 Fee2 10.8958 4.079 .491 .578 Fee3 10.1166 3.471 .573 .508 Fee4 11.2556 3.619 .403 .634 Source: Compiled by the author Coefficient Cronbach’s Alpha = 0.665 is greater than 0.6, so the "Tuition" scale is reliable. Therefore all observed variables in this factor are accepted and continue to be used in EFA factor analysis. Analytical results of the scale "Facilities" Table 413 Facilities Reliability Statistics Cronbach's Alpha N of Items .810 3 Source: Compiled by the author Table 414 Facilities Item-Total Statistics Scale Mean if Item Deleted Scale Variance if Item Deleted Corrected Item-Total Correlation Cronbach's Alpha if Item Deleted Fac1 12.1216 2.923 .353 .646 Fac2 11.8536 2.901 .388 .622 Fac3 11.4491 2.532 .591 .482 Fac4 11.5931 2.739 .430 .594 Source: Compiled by the author The analytical results show that the Coefficient Cronbach’s Alpha = 0.657 is greater than 0.6 than the "Facilities" scale is reasonable and if any observed variable is removed, the Alpha coefficient is less than 0.657. Therefore all observed variables in this factor are accepted and continue to be used in EFA factor analysis. Analytical results of the scale "Recommendations from relatives and friends" Table 415 Recommendations from relatives and friends Reliability Statistics Cronbach's Alpha N of Items .810 3 Source: Compiled by the author Table 416 Recommendations from relatives and friends Item-Total Statistics Scale Mean if Item Deleted Scale Variance if Item Deleted Corrected Item-Total Correlation Cronbach's Alpha if Item Deleted Rec1 8.0744 1.790 .666 .736 Rec2 7.7643 2.036 .718 .692 Rec3 7.7891 1.973 .609 .793 Source: Compiled by the author The Coefficient Cronbach’s Alpha = 0.810 shows that the "Recommendations from relatives and friends" scale is a good scale and if any observed variable is removed, the Alpha coefficient is less than 0.810. Therefore all observed variables in this factor are accepted and continue to be used in EFA factor analysis. Analytical results of the scale "Brand" Table 417 Brands Reliability Statistics Cronbach's Alpha N of Items .646 4 Source: Compiled by the author The analytical results show that the Coefficient Cronbach’s Alpha = 0.646 is greater than 0.6, then the "Branding" scale is appropriate and if any observed variable is removed, the Alpha coefficient is less than 0.646. Therefore all observed variables in this factor are accepted. Table 418 Brand Item-Total Statistics Scale Mean if Item Deleted Scale Variance if Item Deleted Corrected Item-Total Correlation Cronbach's Alpha if Item Deleted Brand1 10.2605 4.472 .478 .539 Brand2 10.2630 5.145 .357 .622 Brand3 10.2779 4.664 .458 .555 Brand4 10.1787 4.635 .411 .588 Source: Compiled by the author Analytical results of the scale “Connections and Bonding in class” Table 419 Connections and Bonding in class Reliability Statistics Cronbach's Alpha N of Items .787 4 Source: Compiled by the author Table 420 Connections and Bonding in class Item-Total Statistics Scale Mean if Item Deleted Scale Variance if Item Deleted Corrected Item-Total Correlation Cronbach's Alpha if Item Deleted Con1 12.4218 3.717 .660 .705 Con2 12.5012 3.340 .611 .731 Con3 12.1489 4.102 .552 .757 Con4 12.2233 3.686 .576 .745 Source: Compiled by the author All observed variables in this factor are accepted because the Coefficient Cronbach’s Alpha = 0.787 and if any observed variable is removed, the Alpha coefficient is less than 0.787. Analytical results of the scale “Decision” Table 421 Decision Reliability Statistics Cronbach's Alpha N of Items .854 6 Source: Compiled by the author Table 422 Decision Item-Total Statistics Scale Mean if Item Deleted Scale Variance if Item Deleted Corrected Item-Total Correlation Cronbach's Alpha if Item Deleted Dec1 18.20 4.511 .640 .829 Dec2 18.15 4.545 .657 .826 Dec3 18.22 4.496 .642 .829 Dec4 18.17 4.578 .635 .830 Dec5 18.15 4.441 .664 .825 Dec6 18.18 4.620 .601 .837 Source: Compiled by the author The "Decision" scale consists of six observed variables. All of these 6 variables have Corrected Item-Total Correlation greater than 0.3 and Cronbach's Alpha if Item Deleted is smaller than Alpha variable - a total of 0.854, therefore they all should be accepted. In addition, Cronbach’s Alpha coefficient is quite high at 0.854 (greater than 0.6) so this scale is satisfactory. These variables are included in the next EFA factor analysis. In summary, by analyzing Cronbach's Alpha for the scales affecting the decision of choosing a foreign language center of students in Ho Chi Minh City, there is only one observed variable on the scale "Geographic location" eliminated. The remaining variables in the standard ranges will be used for further EFA analysis. Exploratory Factor Analysis Exploratory Factor Analysis (EFA) is an analytical technique to minimize and summarize data that is useful for identifying the set of variables needed for a research problem. Using KMO (Kaiser-Meyer-Olkin) and Bartlett test method to measure the compatibility of the sample, if 0.5≤KMO 1 will be retained in the model because these factors work better at summarizing information than an original variable. When conducting factor analysis, the author used Extraction method, Principal Components Analysis, with Rotation Varimax, variables with factor loading factor less than 0.5. is disqualified (Hair & ctg, 1998). According to Gerbing and Anderson (1998), the scale is accepted when the total variance extracted is greater than or equal to 50%. Factor analysis with independent variables Table 423 Factor analysis with independent variables - KMO Kaiser-Meyer-Olkin Measure of Sampling Adequacy. .710 Bartlett's Test of Sphericity Approx. Chi-Square 5120.522 df 666 Sig. .000 Source: Compiled by the author The above result shows that the coefficient KMO = 0.887 (> 0.5) and the significance level Sig = .000 is smaller than the requirement of 0.05 so the observed variables are correlated with each other, so the above factor analysis is suitable. Table 424 Factor analysis with independent variables Total Variance Explained 1 Component Initial Eigenvalues Extraction Sums of Squared Loadings Total % of Variance Cumulative % Total % of Variance Cumulative % 1 5.480 14.810 14.810 5.480 14.810 14.810 2 3.062 8.277 23.087 3.062 8.277 23.087 3 2.267 6.127 29.214 2.267 6.127 29.214 4 2.119 5.726 34.940 2.119 5.726 34.940 5 1.964 5.309 40.249 1.964 5.309 40.249 6 1.797 4.857 45.106 1.797 4.857 45.106 7 1.711 4.624 49.730 1.711 4.624 49.730 8 1.669 4.511 54.241 1.669 4.511 54.241 9 1.324 3.579 57.820 1.324 3.579 57.820 10 1.275 3.445 61.265 1.275 3.445 61.265 Source: Compiled by the author Table 425 Factor analysis with independent variables Total Variance Explained 2 Component Initial Eigenvalues Extraction Sums of Squared Loadings Total % of Variance Cumulative % Total % of Variance Cumulative % 11 1.172 3.169 64.433 1.172 3.169 64.433 12 1.052 2.844 67.277 1.052 2.844 67.277 13 .965 2.607 69.884 14 .857 2.315 72.200 15 .810 2.188 74.388 16 .753 2.035 76.423 17 .690 1.865 78.289 18 .682 1.843 80.131 19 .651 1.760 81.891 20 .565 1.527 83.418 21 .559 1.511 84.928 22 .493 1.333 86.262 23 .488 1.320 87.582 24 .481 1.301 88.882 25 .464 1.253 90.135 26 .438 1.183 91.319 27 .391 1.056 92.375 28 .371 1.003 93.377 29 .346 .934 94.312 30 .341 .921 95.233 31 .330 .891 96.124 Source: Compiled by the author SAU MỖI BẢNG PHẢI CÓ PHẦN PHÂN TÍCH, NHẬN XÉT CHỨ KHÔNG THỂ CHỈ TRÌNH BÀY CÁC BẢNG LIÊN TỤC NHAU THÌ NGƯỜI ĐỌC KHÔNG BIẾT TÁC GIẢ MUỐN DIỄN ĐẠT GÌ! Table 426 Factor analysis with independent variables – Rotated 1 Component 1 2 3 4 5 6 7 8 9 10 11 12 Mkt2 .804 Mkt4 .798 Mkt3 .767 Mkt1 .748 Con1 .785 Con4 .760 Con2 .751 Con3 .711 Pro4 .816 Pro3 .720 Pro5 .647 Pro2 .609 Pro1 .591 Loca2 .846 Loca1 .827 Loca3 .767 Rec2 .866 Rec1 .835 Rec3 .821 Fee2 .751 Fee3 .728 Fee4 .721 Brand1 .725 Brand3 .714 Brand2 .604 Brand4 .583 Source: Compiled by the author Table 427 Factor analysis with independent variables – Rotated 2 Component 1 2 3 4 5 6 7 8 9 10 11 12 Fac3 .763 Fac1 .719 Fac4 .625 Fac2 .559 Qua3 .795 Qua1 .729 Qua2 .562 Tea2 .757 Tea3 .717 Tea1 .618 Fee1 .709 Source: Compiled by the author With the Principal Components extraction method and Varimax rotation, the factor analysis extracted 10 factors from 37 observed variables and with a total variance extracted 67.277% (greater than 50%), components stretch from 0.559 to 0.866 and both are greater than 0.5. Drawn scales are accepted. However, the variable "Fee1" after the rotation matrix is used in its own column, so this observation variable is excluded. (The variable "Fee1" is “Having very high tuition fees and having commitment to refund tuition if students cannot achieve their goals.”) By rotating the factors, the factor matrix will become simpler to explain. We use the Varimax procedure method to rotate the element: rotate the whole angle of the elements to minimize the number of variables with large coefficients at the same factor, thus enhancing the ability to explain the factors (Trong, Ngoc, 2005). Fortunately, after using the rotation matrix, the 10 groups were split back in the same way as the author's scaling. So because these 10 groups will be the 10 new factors to be included in the correlation test and used to run regression, except for Fee1 which is excluded from the Tuition Fees group (explained above). Dependent scale Table 428 Dependent scale - KMO Kaiser-Meyer-Olkin Measure of Sampling Adequacy. .878 Bartlett's Test of Sphericity Approx. Chi-Square 899.562 df 15 Sig. .000 Source: Compiled by the author The above result shows KMO coefficient = 0.878 (> 0.5) and with Sig significance level. = 0.000 <0.05 so the observed variables are correlated, so the analysis of the above factors is perfectly appropriate. Table 429 Dependent scale - Total Variance Explained Initial Eigenvalues Extraction Sums of Squared Loadings Total % of Variance Cumulative % Total % of Variance Cumulative % 3.468 57.796 57.796 3.468 57.796 57.796 .637 10.617 68.413 .564 9.401 77.814 .513 8.555 86.369 .424 7.075 93.444 .393 6.556 100.000 Source: Compiled by the author Table 430 Dependent scale - Component Component 1 Dec5 .782 Dec2 .775 Dec3 .762 Dec1 .760 Dec4 .755 Dec6 .726 Source: Compiled by the author The results of the Decision dependent variable analysis with extracted EFA were collected into one element at Eigenvalue = 3,468 including six observed variables Dec1, Dec2, Dec3, Dec4, Dec5, Dec6 with a KMO index of 0.878. Observed variables all have loading factors greater than 0.50. The variance extracted by 57,796% (> 50%) indicates that these six factors explain the variation of 57,796% of the data. The EFA analysis is completed because it has reached statistical reliability. So the scale is used for further analysis. Test the research model by regression analysis After performing discovery factor analysis, we conduct multiple regression analysis, check VIF (Variance Inflation Factor). If the assumption of multicollinearity is not violated, a multiple linear regression model is built. And the adjusted R2 coefficient (Adjusted R Square) shows how well the regression model is built. To determine the causal relationship between variables in the model, the first step is to analyze the correlation between the variables to see if there is a linear relationship between the independent and dependent variables. Although the results of this analysis do not determine the causal relationship between the dependent and independent variables, it serves as a basis for regression analysis. The highly correlated and independent variables signify the existence of a potential relationship between the two variables. At the same time, the correlation analysis also serves as a basis for detecting the hypothetical violation of linear regression analysis: the independent variables are highly cor