Most studies begin with the hypothesis of the relationship between two variables. For example, based on a certain theory, we may predict that women pay less attention to politics than men, or that social status is positively related to self-confidence. These assumptions almost always predict the relationship between two variables. Therefore, the first step of data analysis is to test whether the relationship predicted by these assumptions exists, that is, to describe the existence, strength and internal situation of this relationship. It answers the question of what is a social phenomenon. The most basic method to describe the relationship between two variables is the "interactive classification" method, also known as contingency wheat. The table15-1is a 3×3 contingency table, which consists of the variable "education level of young people" and the variable "maximum will".

Through interactive classification, the relationship between variables is presented. For example, from the above table, we can see the influence of different educational levels on the maximum volunteers: the maximum volunteers with low educational levels are mostly ideal jobs, while those with high educational levels are happy families and wide knowledge.

The above descriptive analysis statistically points out the existence and size of the relationship between the two variables. However, the last section pointed out that whether the two variables are statistically related is not necessarily completely consistent with whether there is an internal relationship in fact. So this descriptive analysis still can't answer the question whether the relationship between the two variables predicted by the fake mother really exists.

In addition, this descriptive analysis can't answer questions such as "why is this relationship" and "under what conditions or circumstances does this relationship exist". Answering these questions is the task of the second step of analysis, that is, to make an accurate causal analysis of the relationship and degree of connection between variables, so as to judge the authenticity of the relationship, answer why this relationship occurs, and explain the conditions for its existence. In order to explain and test the true relationship between two variables, although we can guess according to the existing knowledge, the more valuable method is to carry out systematic test.

Second, the introduction of investigation factors

The most important and systematic way to test the relationship between two variables is to introduce a third variable. Then check the change of the original relationship between the independent variable and the dependent variable after the introduction of the third variable, so as to clarify and deepen the understanding of the original relationship and reveal the real relationship between the two variables. This process of introducing a third variable to test the relationship between two variables to explain or determine the relationship between variables is called detailed analysis, and the introduced variables are called test factors or control variables.

The detailed analysis model was developed by American sociologist Paul lazarsfeld and his assistant, but its main idea came from Samuel Stover's work in his famous book The American Soldier. American Soldier is the result of Christopher's investigation on the morale of American soldiers during the Second World War. The war-weariness of American soldiers is well known, so what are the factors that produce this mood or affect the morale of the army? He first tested some accepted assumptions, such as: "The more educated people are, the less willing they are to join the army." Wait a minute. But surprisingly, the survey results are contrary to these accepted patterns, for example, the less educated people are, the less willing they are to join the army.

What is the reason? Christopher found the answers to these results in the thoughts of reference group and relative deprivation. To put it simply, he believes that soldiers evaluate their situation not by absolutely objective standards, but by comparing themselves with those around them. When people find themselves "disadvantaged" compared with the people around them (that is, his reference group). He will have a sense of relative deprivation, that is, he seems to feel that he has been deprived of something by others. Based on the reference group theory and relative deprivation theory, this paper explains why people with low education are more reluctant to join the army: people with low education also have friends with them. In wartime, because people with low education are more engaged in national defense industry or national defense production, more people are exempted from enlistment, so people who join the army feel particularly disadvantaged compared with their friends. These situations do not exist among people with high academic qualifications.

Stoffer's explanation makes the survey results reasonable, but because these situations were not expected during the research and design, it is impossible to confirm the above explanation with empirical data, but his logical explanation paves the way for the establishment of a detailed analysis model. Mainly through other variables (reference group-friends) to explain the relationship between the two variables (education level and willingness to apply). Christopher's work was later continued by lazarsfeld and his colleagues, who confirmed Christopher's explanation with data and developed a detailed analysis model. We use an example to illustrate how to use the detailed analysis model, that is, how to use the test factor to test the relationship between two variables.

When forced to ask why this phenomenon occurs, the researchers assume that it is influenced by the education level, that is, older people may like to listen to religious programs because of their low education level. In order to test this hypothesis, the respondents were grouped according to their different education levels and made into a table 15-3.

In Table 15-3, the proportion of young people and old people listening to religious programs is 2% (1 1%-9%), and it is 3% (32%-29%) in the low-education group, both of which are far less than 9%. This shows that when the influence of education is eliminated, the proportion of young people and old people listening to religious programs is very small. In other words, age has nothing to do with listening to religious programs, but the original relationship between them is caused by education, because both are related to education at the same time. This answers the question of the authenticity of the relationship between two variables and "why is there such a relationship?" Old people prefer to listen to religious programs because of their low education level, while people with low education level prefer to listen to religious programs.

In this example, the variable as the test factor is the education level, and the so-called "split table method" is used in the investigation process. Specifically:

1. First, describe the relationship between variables x and y (table 15-2 in this example). The relationship at this time is called the original relationship.

Choose test factors according to theory or experience (in this case, education level).

3. Divide the test factors into different grades or categories (in this case, it is divided into high-education group and low-education group), and then make a table of X and Y in each category (in this example, table 15-3 includes two separate tables of high-education group and low-education group). The relationship between x and y in the sub-table is called partial relationship.

4. Investigate the relationship between X and Y in each table (that is, the partial relationship).

(1) If the original relationship between X and Y disappears in all sub-tables (that is, X and Y are irrelevant in all sub-tables), it is proved that the original relationship is mainly caused by test factors;

(2) If the original relationship between X and Y still exists in each sub-table (that is, the relationship between X and Y in each sub-table is similar to the original table), it means that the relationship between X and Y is not affected by the test factors;

(3) If the original relationship between X and Y exists in each sub-table, but it is weaker than the original relationship, it proves that the relationship between X and Y is partly influenced by the test factors;

(4) If the original relationship between X and Y exists or strengthens in some sub-tables, but disappears or weakens in others, it means that the existence of the original relationship between X and Y is conditional.

The first three cases are called general relations, and the last case is called conditional relations.

Thirdly, the main functions of the model are analyzed in detail.

One of the main functions of detailed analysis model is to share some advantages of experimental design in investigation and research. Besides deduction of mathematical logic, experiment is the most powerful proof model in scientific research. Its theoretical basis is the so-called "difference method", that is, if one example appears in the phenomenon under investigation and the other example does not appear, the two examples are the same except one point, then what makes the two examples different is the reason for the phenomenon. Therefore, we can choose two identical groups to compare, only give one group some kind of stimulation, and then observe whether the two groups are different. If there are differences, this stimulus is the reason, which is the characteristic of "post-experimental design". In the study of social phenomena, due to various reasons, it is often impossible to carry out direct experiments, but only in an indirect way. The detailed analysis model is similar to the post-experimental design.

For example, case 1, if you want to know whether age is the reason for different interests in listening to religious programs, you must find two groups according to the principle of experimental methods; They are all the same except age, and then compare their listening interests. However, it is impossible to find the same two groups, so it controls irrelevant factors, such as education level, and narrows the differences between the two groups through variable control. If these unrelated factors are controlled, there are still differences in listening interest between the two groups of different ages, then it can be safely said that age is a reason. There is no doubt that the more items are controlled, the closer the two groups may be to the same except for one variable. In this way, the detailed analysis model makes the investigation similar to the post-experimental design, thus becoming one of the most powerful proof models in social science.

Detailed analysis of the model can make full use of statistical survey data and make the research deeper. On the one hand, it can describe the relationship between variables, on the other hand, it can clarify the truth by introducing a third variable. These facts include the truth value of the relationship between two variables or the conditions and reasons for the existence of this relationship. Therefore, the relationship between variables is more specific, accurate and reliable. The purpose of analysis is to explain, and the detailed analysis model has great contribution to the explanation. It can not only confirm and help explain, but also eliminate the wrong explanation and get a new one. Therefore, it is a powerful tool for establishing theories and developing materials.

The detailed analysis model provides people with the clearest social analysis logic. Only by understanding and mastering the ideas can we understand more complex social statistics skills.