This article mainly introduces the concept of mediation effect analysis and path analysis, as well as operation steps and precautions. A lot of the content is the copy of Chapter 12 of the "Quantitative Research and Statistical Analysis: SPSS (PASW) Data Analysis Example Analysis" by Qiu Haozheng, and then added some understanding of the process that I saw in the middle.
1. Analysis of intermediary effects
From the above explanation, we can know that the mediating effect analysis actually has a variable acting as the mediating variable, and X->Y transmits the influence through the mediating variable Z. If it is a complete mediation effect, it is completely transmitted by Z. If it is a partial mediation effect, it is partially transmitted by Z and partially directly affected.
2. Path analysis
Path analysis actually feels more than a method of innovation. It just piles up the mediation effect analysis, and then performs a regression on each endogenous variable to see if the regression coefficient is significant, and then follows the same as the mediation effect analysis. The process analyzes the direct effects and indirect effects. Finally, the effect of a certain variable X on another variable Y is the sum of all direct effects and indirect effects.
But here I have some doubts. There is such a sentence in the book (see the figure below). I think the insignificant ones don’t need to be summed up. Of course, this effect is very small, because if it is not significant, the coefficient is naturally Tend to 0, the effect of adding and not adding up is not big, but from the actual meaning, I personally prefer to not adding up, because not significant means that there is no correlation, so it should be considered that there is no such impact. It should not be added up.
3. Matters needing attention
There is not much to pay attention to in the analysis of feeling mediation effect and path analysis. This is more of a kind of thought. The method is actually ordinary multiple regression, but note that the path coefficient is the standardized coefficient, not the original coefficient. Then if you use spss, the significance of the mediation effect seems to need to be tested by yourself, which is a little troublesome, and the others seem to be fine.
4. References
"Quantitative Research and Statistical Analysis: Analysis of SPSS (PASW) Data Analysis Example" by Qiu Haozheng
——————————————————————————————————
Because there is no engineering requirement now, in fact I have not really run through the intermediary Effect analysis and path analysis, and then add your own practical experience.