CHAPTER 1: THEORETICAL BASIS OF THE LOYALTY OF STAFF AND
2.9.1. Regression model for component variables
Regression analysis was carried out with 6 independent variables including:
Colleagues (X1), Leadership (X2), Training and promotion (X3), Salary (X4), Nature of work (X5), The working environment (X6) and the dependent variable are Employee Loyalty (Y).
The values of the independent variables are averaged based on the component observed variables of those independent variables. The value of the dependent variable is the average value of the observed variables of employee loyalty. Analysis is done by the Enter method. Variables are included at the same time to see which variables are accepted. The results of regression analysis are as follows:
Ingredient
Not standardized Beta stand ardiz ed
Value t
Value P
Measurement of multicollinearity (Collinearity Statistics)
B False
Acceptance (Tolerance)
Magnification coefficient variance (VIF) Constant
-.206 .281 -.732 .465
Colleague
.093 .058 .080 1.616 .107 .769 1.301
Leader
.300 .069 .249 4.337 .000 .570 1.755
Training and
promotion .074 .066 .066 1.121 .264 .542 1.846
Salary
.228 .067 .181 3.387 .001 .654 1.529
The nature of the work
.421 .057 .400 7.420 .000 .645 1.551
Work environment
-.045 .041 -.048 -1.098 .273 .961 1.041
R2 calibrated = 0.558
Table 10 Regression results using the Enter method
The results of multiple linear regression analysis with the enter method (a one-time input method) showed that the model with R2 adjusted to reach 0.558 is shown (Table 3.23).
From the results of the above regression model, we see that the factors of working, training and advancement, colleagues are not statistically significant, P value is greater than 0.05.
Assessing the appropriateness of the model: Through Adjusted R square to assess the suitability of the model will be safe because it does not inflate the relevance of the model.
Model with R2 adjusted is 0.558. This speaks of the model's appropriateness of 55.8% or in other words this model explains 55.8% of the variation of Fidelity due to variables in the model and 44.2% of the variation of the Fidelity factor is explained by variables other than the model that in the scope of this study has not been considered.
Ingredient Coefficient R R 2 R2 calibrated Standard errors of estimates
1 .755 a .570 .558 .58289
Table 11 Evaluation of conformity of the model
a. Independent variables: (Constant), Working environment, Colleagues, Salary, Leaders, Work nature, training and advancement
Testing the suitability of the model: Test F is used to consider the relationship between the dependent variable Y that has a linear relationship with the whole set of variables. If the hypothesis R2pop = 0 is rejected, the combination of the existing variables in the model can explain the change of the Y variable, ie the construction model is suitable for the data set.
In ANOVA analysis table, we see the value of sig. very small (sig = 0.00
<0.05), rejecting the R2pop hypothesis = 0 or concluding that the regression model is suitable for the overall and usable. ANOVA analysis for value F = 50,722
Ingredient Total squared
deviations
Degrees of freedom
Average squared deviation
F Sig.
1 Regression 103.402 6 17.234 50.722 .000 b
Residuals Total
78.146 181.548
230 236
.340
Table 12 Model conformity test table
a. Dependent variable: Loyalty b. Independent variables:: (Constant), Working environment, Colleagues, Compensation, Leadership, Work nature, training and advancement
Detecting violations of assumptions in linear regression:
Assumption of linear relationship: Using a scatter plot between residuals and predicted values given by the linear model.
Through the graph below, it can be seen that this assumption is rejected because the normalized balance disperses quite randomly on both sides of the line through 0.
Assume the variance of constant error: Use the dispersion graph between the residuals and the predicted values given by the model, if the magnitude of the residuals increases or decreases with predicted values, it can be suspected I suspect that the variance of the constant error is violated. Besides using Spearman test, with the hypothesis set to be the variance of the variance error, if this hypothesis is correct, the overall rank correlation coefficient between the remainder and the independent variable will be different from zero.
Considering the above graph, we can reject the assumption that the variance changes because the normalized residual disperses quite randomly. To ensure that the study has performed Spearman test for the correlation between the independent variables and the residual with the hypothesis H0 is the correlation coefficient of the whole is 0
Correlations
ABSRES Wor
Correlation coefficients 1.000 -.099
ABSRES Sig. (2-tailed) . .130
Sample size 237 237
Spearman's rho Correlation coefficients -.099 1.000
Wor Sig. (2-tailed) .130 .
Sample size 237 237
Figure 13 Scatter chart between residuals and predicted values
Correlations
ABSRES Sup
Correlation coefficients 1.000 -.119
ABSRES Sig. (2-tailed) . .067
Sample size 237 237
Spearman's rho Correlation coefficients -.119 1.000
Sup Sig. (2-tailed) .067 .
Sample size 237 237
Correlations
ABSRES Pay
Correlation coefficients 1.000 -.062
ABSRES Sig. (2-tailed) . .344
Sample size 237 237
Spearman's rho Correlation coefficients -.062 1.000
Pay Sig. (2-tailed) .344 .
Sample size 237 237
Correlations
Cow ABSRES
Correlation Coefficient 1.000 -.053
Cow Sig. (2-tailed) . .415
Spearman's rho N 237 237
Correlation Coefficient -.053 1.000 ABSRES
Sig. (2-tailed) .415 .
N 237 237
Correlations
Pro ABSRES
Correlation Coefficient 1.000 -.172 **
Pro Sig. (2-tailed) . .088
N 237 237
Spearman's rho -.172 ** 1.000
Correlation Coefficient
ABSRES Sig. (2-tailed) .008 .
N 237 237
Correlations
Env ABSRES
Spearman's rho
Env
ABSRES
Correlation Coefficient Sig.(2-tailed)
N
Correlation Coefficient Sig.(2-tailed)
N
1.000 -.014
. .833
237 237
-.014 1.000
.833 .
237 237
Table 13 Spearman test
Through the above tables, sig. respectively 0.130, 0.067, 0.344.0.415, 0.088, 0.833 are all larger than 0.05, thus accepting the H0 hypothesis. From this, it is concluded that the variance of the error is constant.
Assumptions about the normal distribution of remainder: Use the Histogram frequency chart, or the plot plot of the QQ plot, or the PP plot frequency chart, if the dots are dispersed close to the diagonal, redistribution can See as standard.
Through three graphs: Histogram, P-P plot, and Q-Q plot, acceptable is the standard distribution balance.
Figure 14 Frequency chart of standardized residuals
Assumptions about the independence of the error (there is no correlation between the residuals): the assumption of true error ei is a random, independent variable, with a normal distribution with a mean of zero and a variance change.
Using the Durbin - Watson (d) statistical quantity with the hypothesis set is: the overall correlation coefficient of the remainder is 0. If the remainder has no first- order chain correlation, the value d will almost equal to 2.
Durbin-Watson d's salary statistics are worth 2,032 roughly equal to 2, so it can be determined that there is no string correlation between the remainder.
Model Summary b Ingredi
ent R coefficient
R 2 R2 calibrated Standard errors of estimates
Durbin-Watson
1 .755 a .570 .558 .58289 2.032
Table 14 Durbin Watson statistics
a. Independent variable: (Constant), Env, Cow, Pay, Sup, Wor, Pro b. Dependent variable: Loy
Multicollinearity detection: The tools for detecting the existence of linear additions in the data and assessing the level of collinearity are degradation of the estimated parameters: Acceptance of the variable (Tolerance) - if tolerance of a small variable, there may be a linear combination of other variables that explain the independent variable. Magnification coefficient of variance (Variance inflation
Figure 16 P-P Plot frequency chart of standardized residuals
factor - VIF), when VIF exceeds 10 it is a sign of multicollinearity.
In the case of independent variables having multicollinearity phenomenon, ie independent variables are closely correlated with each other. It provides very similar information to the model, making it difficult to separate the effects of individual variables. In order to avoid misinterpreting regression results, it is necessary to assess and measure multicollinearity phenomenon. With great acceptance (Tolerance). The magnification coefficient of variance variance (VIF = Variance inflation factor) from 1,041 to 1,846 is less than 10, so the relationship between these independent variables is negligible. There is no multicollinearity phenomenon. It is safe to use the regression equation. Values of VIF = 1 / Tolerance (Trong and Ngoc, 2008).
Thus, the linear regression model is constructed without violating the necessary assumptions in linear regression.
The results in Table 3.23, the regression equation show the relationship between Employee Loyalty (Y) and the independent variables expressed by the following expression.
Y = 0.400 * X5 + 0.249 * X2 + 0.181 * X4 + 0.080 * X1 + 0.066 * X3 + 0.048 * X6 (Sig = 0.0) (Sig = 0.0) (Sig = 0.001) (Sig = 0.107) (Sig = 0.264) (Sig = 0.273) Inside:
Y: Loyalty X1: Colleagues X2: Leadership
X3: Training and promotion X4: Salary
X5: Nature of work
X6: Working environment
Regression results show independent variables Work nature, Leadership, Salary with Sig. less than 0.05 so these variables are significant at the 95%
confidence level. Therefore, at the 95% confidence level of these variables affecting Loyalty and the slope coefficients of 0.40, 0.249, 0.181 respectively, they
are positive, so the variables are in the same direction as the Loyalty of employees.
The importance of the variables of Leadership, Salary, Nature of work for Loyalty variable is determined based on Beta coefficient. If the absolute value of the Beta coefficient of any factor is greater, the more important it is to affect employee loyalty to the company. Therefore, the most important influence on employee loyalty is the nature of work (Beta = 0.40), followed by Leadership (Beta
= 0.249) and finally the Wage factor (Beta = 0.181).
Thus, based on the regression results, there are three factors that affect the loyalty of employees at ANSV Co., Ltd.: Leadership, Salary, Nature of Work. In particular, the element of work nature has the most impact on employee loyalty.
For factors excluded from the model: when interviewing directly in the qualitative research step, the staff responded that this factor is influential. However, the results of quantitative analysis according to the linear regression model show that these factors are not suitable. This can be explained as follows: The majority of employees who leave the interviewed company think that colleagues and the working environment at the company are quite good, the policy of training is quite complete. However, they left the company mainly because of low wages, inadequate and timely assessment of the work, leaders treated them unfairly, not empowered to actively handle the work. Therefore, employees who do not see colleagues, work environments, training and promotion are important factors for them to leave the company.
2.9.2. Regression model with the participation of qualitative variables
In the study, the impact of qualitative variables including gender, age, education level, and working time on the relationship in the regression equation was examined.
According to the survey form shown in Table 3.2, employees in the company with a university degree mainly account for 61.18% so we create 1 Edu dummy with 0 being a university level, 1 is a different university level.
Similarly create 3 dummy variables for the age group (age group 1 <= 30, age group 2 from 31-40, age group 3> 40).
Create a sex dummy variable Sex with 0 being male, 1 being female.
Creating a dummy variable for seniority of Sen with 0 is a group with seniority of less than 10 years and 1 is a senior group over 10 years.
The author puts dummy variables of qualitative variables into the same regression model. The results (Table 3.28) show that all qualitative variables are not statistically significant and excluded from the model because there are P values> 0.05. All qualitative variables do not affect loyalty.
Ingredient
Not standardized Beta stand ardiz ed
Valuet Value P.
Measurement of multicollinearity (Collinearity Statistics)
B False
Acceptance (Tolerance)
Magnification coefficient variance (VIF 5 ) Constant
.184 .354 .520 .604
Colleague
.108 .059 .092 1.836 .068 .712 1.405
Leader
.283 .069 .235 4.092 .000 .547 1.827
Training and
promotion .067 .066 .060 1.027 .305 .526 1.901
Salary
.167 .069 .132 2.427 .016 .606 1.650
The nature of the
work .421 .059 .400 7.107 .000 .570 1.755
Work environment
-.049 .042 -.052 -1.178 .240 .916 1.091
Sex
-.130 .088 -.070 -1.479 .140 .805 1.243
Age group 2
-.001 .138 -.001 -.008 .993 .366 2.729
Age group 3
-.238 .181 -.103 -1.316 .189 .295 3.385
Academic level
.067 .084 .037 .799 .425 .828 1.208
Long-term work
.055 .096 .031 .574 .567 .606 1.650
R2 calibrated = 0.575
Summary chapter 2
In chapter 2 of the author's essay, a general overview of ANSV was presented. At the same time, analyze the status of employee loyalty at ANSV Co., Ltd. and provide employee loyalty assessments at ANSV Co., Ltd. On that basis, chapter 3 of the author's essay will go into more detailed analysis, thereby giving suggestions to maintain and enhance employee loyalty at ANSV Co., Ltd.
CHAPTER 3: SUGGESTING SOLUTIONS TO MAINTAIN AND IMPROVE THE LOYALTY OF STAFF AT THE ANSV COMPANY