If a survey is conducted to ask which statistical technology is cutting-edge and most popular in China's current psychological research, the answer must be intermediary analysis. Because there are ready-made statistical software to do this analysis, undergraduates majoring in psychology will consider intermediary testing for future data analysis when they conceive and design research plans.

If you ask a question: "why do you do intermediary analysis?" researchers are likely to say: "intermediary analysis can investigate the psychological mechanism." or "intermediary analysis can investigate the internal mechanism." if basic research is carried out, investigating the internal mechanism is a good reason and goal. However, almost all psychology undergraduates in China specialize in Applied Psychology. Their research, especially for writing graduation thesis, should have been applied research, not basic research. Similarly, graduate students of master of Applied Psychology and master of Education (direction of mental health education) basically have to do intermediary analysis and carry out theoretical research rather than applied research.

In fact, it can be speculated that many researchers conduct intermediary analysis and imitate more components and less for practical needs. In particular, can intermediary analysis really examine the internal mechanism? If we look at the research on intermediary analysis, the problem will be clearer, that is, most of the research using intermediary analysis, the method of collecting data is measurement, and the relevant research is carried out, not experimental research, so it is difficult to investigate the psychological mechanism or internal mechanism. Obviously, about intermediary analysis, we still need to introduce some background knowledge from the perspective of research methods.

The scientific research of psychology begins with investigating the relationship between two variables. If it is an experimental study, one variable x is the independent variable and the other variable y is the dependent variable, the problem is how the dependent variable y changes with the change of independent variable x. In case of relevant research, although it is inconvenient to say that one variable is an independent variable and the other variable is a dependent variable, the problem or intention of the research is to investigate how one variable changes with the change of another variable - only to find the relationship between the two variables, which is not particularly useful. After all, the task of psychology is not only to describe and predict psychological phenomena, but also to explain and control psychological phenomena.

To explain psychological phenomena, it is necessary and inevitable to put forward theories. Controlling psychological phenomena also needs and should be based on theory. Thus, the variables investigated in psychological research may no longer be only independent variable x and dependent variable y, but with variable Z other than these two variables, they are collectively referred to as the third variable. One of the variables m is very important for constructing theory. They are very helpful to clarify the causal relationship between independent variable x and dependent variable y. This kind of variable m is an intermediary variable.

In a psychological study, there can be more than one independent variable, dependent variable and intermediary variable. If a study only involves an independent variable x, a dependent variable y and an intermediary variable m, the corresponding intermediary analysis is the simplest and most basic. At this time, it is assumed that a cause variable (independent variable) affects an intermediate variable, and this intermediate variable affects a result variable (dependent variable). In other words, intermediary variable m is a theoretical concept that transfers the effect of one variable to another variable. It can be behavioral, biological, psychological and social.

Mediation analysis is a method by which researchers explain the process or mechanism by which one variable x affects another variable y. Generally, the research of mediation analysis can be roughly divided into two categories, or mediation analysis plays an important role in two aspects.

The first category examines how a specific effect is produced. This kind of research is an intensive research after discovering the relationship between X → y, that is, in order to better understand the relationship between X → y or test the authenticity of this relationship, a third variable m is added to the analysis of X → y. A third variable can promote the understanding of this relationship because it is the organizational part of the X → m → y causal series. For example, physical abuse in early childhood is associated with subsequent violence. One explanation for this relationship is that children who suffer physical abuse will acquire abnormal patterns of processing social information, which will lead to later violence. The empirical study found evidence of this theoretical intermediary process, that is, the measurement indicators of social information processing can explain the relationship between physical abuse in early childhood and subsequent violence.

The second is to design experiments with the theory of intermediary process. A common example of such studies is the evaluation of treatment and prevention programmes. In such a study, an intervention is designed to change the intermediary variable, assuming that there is a causal relationship between the intermediary variable and the dependent variable. If the assumed relationship is correct, a prevention or treatment program can substantially change the mediating variable and the outcome variable. It can be seen that intermediary analysis provides evidence of why a scheme is effective.

In particular, the mediation effect test has some important premises. For example, there is no interaction between the independent variable x and the intermediary variable m, the setting of the intermediary model is correct (that is, X → m → y, not y → m → x), and the intermediary variable m and the dependent variable m are not mutually causal. In addition, it is also a very important variable that does not affect the independent variable x, dependent variable y and intermediary variable M. These premises may be difficult to test and in most cases impossible to test. Therefore, it is impossible to prove an intermediary relationship. A more realistic approach is to combine the existing research information, especially the existing theories, to support the proposed mediation model.

In short, mediation analysis is much more complex than most researchers think, and it is usually difficult to investigate the internal mechanism. For undergraduates and postgraduates engaged in applied psychology research, if they are not sure that the selected intermediary variables may change easily, so that the research results are difficult or unable to be applied, they should not design and implement the research scheme using intermediary analysis.