The role of decentralization of mediating effect variables
Publish: 2021-04-21 16:06:09
1. According to Hou Jietai: the so-called centralization refers to subtracting the mean value of a variable from its expected value. For sample data, each observation value of a variable is subtracted from the sample average value of the variable, and the transformed variable is centralized
for your question, subtract the mean from each measurement.
for your question, subtract the mean from each measurement.
2. 1. The mediating effect analysis does not need data centralization and standardization
2. Forced centralization or centralization, only the non standardization coefficient is different, the standardization system is the same
(provided by Nanxin)
2. Forced centralization or centralization, only the non standardization coefficient is different, the standardization system is the same
(provided by Nanxin)
3. The mediating and moderating effects can be realized by hierarchical regression in SPSS, that is, in which dialog box of multiple linear regression analysis, there is a
Block dialog box, you can move the independent variables and moderating variables to which dialog box one by one, and the regression results will show the changes of moderating effects
Block dialog box, you can move the independent variables and moderating variables to which dialog box one by one, and the regression results will show the changes of moderating effects
4. Yes, because it can be verified in two times. Each regression model considers the relationship between independent variables, intermediate variables and dependent variables when they interact with each other. Therefore, when only these three variables work at the same time, the intermediate variables play a full mediating role. You're not thinking about two intermediate variables working at the same time. But of course, this has its limitations. It is best to use structural equation model to verify.
5. Your model design is too complex. It's better to simplify it properly, especially on the action points of control variables Intermediary moderating effect analysts (nanxinwang can help you)
6. Many people are asking, when is the mediating effect, what is the moderating effect, and what is the interaction? Next, brother song uses three pictures to explain the three concepts, hoping to solve the puzzle
the figure above shows the mediating effect pattern. The effect of a on C occurs through B, that is, a-b-c. If the effect of a-C is zero, then B is a complete mediator; If the effect of a-C is not zero, then B is a partial mediator. Image metaphor: the intermediary effect is "matchmaker", and a-c's understanding is through matchmaker
the above figure shows the regulatory effect. A-c has an effect, but B will affect the effect of a-c. Figurative metaphor, the moderating effect of "small three", will affect a-c normal husband wife relationship
the figure above shows the pattern of I frontal interaction, a-c is related, B-C is related; And B will affect the a-c relationship, a will affect the B-C relationship. It's just like a and B are roommates in the same dormitory. They both like C at the same time, which means AB is junior to each other, but it doesn't matter in sequence
when the forest is big, there are all kinds of birds; Similarly, when we study more factors, everything will happen. When the matchmaker comes, the third child arrives. Sometimes they are the third child, sometimes the matchmaker leads the third child, sometimes the third child takes a fancy to the matchmaker. The situation is not optimistic, but many relationships are composed of matchmaker and junior three
there is no essential difference between regulatory effect and interaction effect in statistical model; But the regulatory effect can specify who is the independent variable and who is the regulatory variable; The interaction status is equivalent
to study mediating and moderating effects, but process is the best choice when the research factors are significant variables; Amos is better when it is a latent variable. Of course, LISREL, Mplus, etc. LISREL is the earliest structural equation modeling software, which has been graally replaced by programming
the figure above shows the mediating effect pattern. The effect of a on C occurs through B, that is, a-b-c. If the effect of a-C is zero, then B is a complete mediator; If the effect of a-C is not zero, then B is a partial mediator. Image metaphor: the intermediary effect is "matchmaker", and a-c's understanding is through matchmaker
the above figure shows the regulatory effect. A-c has an effect, but B will affect the effect of a-c. Figurative metaphor, the moderating effect of "small three", will affect a-c normal husband wife relationship
the figure above shows the pattern of I frontal interaction, a-c is related, B-C is related; And B will affect the a-c relationship, a will affect the B-C relationship. It's just like a and B are roommates in the same dormitory. They both like C at the same time, which means AB is junior to each other, but it doesn't matter in sequence
when the forest is big, there are all kinds of birds; Similarly, when we study more factors, everything will happen. When the matchmaker comes, the third child arrives. Sometimes they are the third child, sometimes the matchmaker leads the third child, sometimes the third child takes a fancy to the matchmaker. The situation is not optimistic, but many relationships are composed of matchmaker and junior three
there is no essential difference between regulatory effect and interaction effect in statistical model; But the regulatory effect can specify who is the independent variable and who is the regulatory variable; The interaction status is equivalent
to study mediating and moderating effects, but process is the best choice when the research factors are significant variables; Amos is better when it is a latent variable. Of course, LISREL, Mplus, etc. LISREL is the earliest structural equation modeling software, which has been graally replaced by programming
7. As for whether D is a mediating variable, on the one hand, you need to find the support of literature. For example, previous studies have mentioned that D may be a mediating variable, but it has yet to be verified, or someone has verified the mediating effect of D. At this point, you can do it again. On the other hand, if there is no research on the mediating effect of D, but logically, the D variable may play a mediating role, you can also try to analyze it. In short, you can analyze it first, and it may not be shown in the article at last
when analyzing the mediating effect, I think it should be done from the level to the whole. Firstly, the mediators of B were analyzed according to a1-b-c, a2-b-c, a-b-c, d1-b-c, d2-b-c and d-b-c. Analyze the mediators of D, a1-d1-c, a2-d1-c, a-d1-c, a1-d2-c, a2-d2-c, a-d2-c, a1-d-c, a2-d-c, a-d-c. Don't be too troublesome, or that sentence, you can first analyze, and finally not necessarily show in the article. We should fully explore the data, so that we can find more things
the analysis method of specific mediating effect, taking a-b-c as an example. First, establish a regression equation with a as the independent variable and C as the dependent variable to verify whether the coefficient of a is significant. Since you have identified a as the independent variable and C as the dependent variable, then the coefficient must be significant. Then, the regression equation with a as the independent variable and B as the dependent variable is established to verify whether the coefficient of a is significant. Finally, we establish a and B as independent variables and C as dependent variables, and look at the coefficient of a and B. if the coefficient of B is significant and the coefficient of a is not significant, it is a complete intermediary; If a and B coefficients are significant, they are partial mediators
the above is a relatively common three-step verification intermediary, and the regression analysis of SPSS can be used. If you want to use structural equation model, take Amos as an example, draw two variable AC model and three variable ABC model, and do path analysis respectively to see if the path coefficient is significant. In fact, the principle is the same.
when analyzing the mediating effect, I think it should be done from the level to the whole. Firstly, the mediators of B were analyzed according to a1-b-c, a2-b-c, a-b-c, d1-b-c, d2-b-c and d-b-c. Analyze the mediators of D, a1-d1-c, a2-d1-c, a-d1-c, a1-d2-c, a2-d2-c, a-d2-c, a1-d-c, a2-d-c, a-d-c. Don't be too troublesome, or that sentence, you can first analyze, and finally not necessarily show in the article. We should fully explore the data, so that we can find more things
the analysis method of specific mediating effect, taking a-b-c as an example. First, establish a regression equation with a as the independent variable and C as the dependent variable to verify whether the coefficient of a is significant. Since you have identified a as the independent variable and C as the dependent variable, then the coefficient must be significant. Then, the regression equation with a as the independent variable and B as the dependent variable is established to verify whether the coefficient of a is significant. Finally, we establish a and B as independent variables and C as dependent variables, and look at the coefficient of a and B. if the coefficient of B is significant and the coefficient of a is not significant, it is a complete intermediary; If a and B coefficients are significant, they are partial mediators
the above is a relatively common three-step verification intermediary, and the regression analysis of SPSS can be used. If you want to use structural equation model, take Amos as an example, draw two variable AC model and three variable ABC model, and do path analysis respectively to see if the path coefficient is significant. In fact, the principle is the same.
Hot content