When carrying out research, especially when writing papers, Chinese psychological researchers often encounter a big problem, that is, the sampling of research objects. No matter how much effort researchers make to improve the representativeness of the sample, there are often reviewers or reviewers who question the representativeness of the research results. Some researchers wrote in the paper: "this study randomly selected XXX subjects." If a reviewer or reviewer asks about the "random selection" program, then judging from the sampling process described by the researcher, the researcher absolutely does not take random samples, and the samples obtained are not probability samples at all.

Thus, psychological research is stuck in the sample, many researchers feel headache, many reviewers and reviewers hold on.

In fact, psychological researchers and reviewers do not have to care too much about the samples, especially the researchers, and they do not have to hide their own sampling process and sample situation. It is convenient to sample and describe honestly. The practice of forcibly writing "random selection" is counterproductive - it is either ignorant and fearless, or deliberately deceptive.

Starting from the methodology of psychological research, I make two interpretations of psychological research, hoping that researchers and reviewers will no longer be entangled in the issue of sampling.

Let's start with some preliminary knowledge: psychology is the science of studying human behavior and consciousness; in the research of psychology, behavior and consciousness should be abstracted as variables; psychological research generally studies the whole through samples - although the whole of the research may depend on the situation; psychological research should be carried out by measuring variables; psychological research is reflected in the investigation of variables The relationship between, especially causality.

From the above knowledge, two important points can be drawn, that is, psychological research needs to measure variables, and psychological research focuses on the causal relationship between variables.

There is an important theory in psychometrics, which is called the classical test theory. Its content is very simple, but very profound, that is, the test score is equal to the sum of the true score and the measurement error. Test scores are the scores of specific measurement, true scores are the psychological attributes (such as intelligence and personality) that researchers are really interested in, and measurement errors are the deviations in the measurement process. In order to make the test scores better express the psychological attributes, it is necessary to control or reduce the measurement error.

The measurement error includes random error and systematic error. The systematic error is mainly affected by the research conditions (for example, the researchers use a set of intelligence scale without revision, which leads to the low IQ). The researchers can identify through the relevant theory or reality, so as to control or reduce it. Random error is caused by random factors in the process of measurement. It is difficult for researchers to identify it. The method to control or reduce it is to make multiple measurements. Theoretically, when the number of measurements approaches infinity, the random error is equal to 0.

Classical test theory only considers random error, not systematic error. (it can be considered that the systematic error is a special random error - if enough measurements are made, the systematic error caused by the research conditions will be equal to zero.) In order to reduce the measurement error, it is necessary to measure the psychological attributes of the research many times. However, it is very difficult to measure the same test repeatedly. For example, N copies of the test are needed, and the subjects are tired of the same test and do not cooperate. The solution that psychometrists think of is to measure many subjects only once. They proved mathematically that in terms of a certain psychological attribute, measuring n times for one person is equivalent to measuring n times for one person. This is the principle of large-scale psychological measurement, although many people who use psychological measurement methods to carry out research do not really understand and understand.

It is not difficult to see that measuring one person n times is equal to measuring one person n times. That is to say, for the psychological attributes that researchers are interested in, measuring one person n times can achieve the effect of measuring n people. If n is equal to the population, then n measurements of one person are equal to the measurement of the population. (of course, theoretically, the n of this should be large enough.) In other words, the measurement of one person can achieve the overall effect of measurement. Obviously, it is also possible to study the population by measuring a sample with a capacity larger than 1. In this case, what's wrong with researchers using a sample to carry out psychological research?

In addition, the current meta-analysis technology, that is, macro analysis of the results of many similar studies, can theoretically solve the error caused by the research sample deviation.

Psychological research focuses on the causal relationship between variables, which can be interpreted from the validity of psychological research. Generally, the validity of psychological research can be divided into four kinds: internal validity, structural validity, statistical validity and external validity. Structural validity corresponds to the theoretical level of the research topic, especially the agreement between the theoretical concept and the measurement index. Statistical validity involves statistical correctness in research. According to my understanding, these two kinds of validity are the basic requirements of psychological research. The following is devoted to internal validity and external validity.

For psychological research, internal validity is the degree of truth of causality between the argument and the dependent variable, while external validity is the degree of truth of the conclusion in other situations. In a study, the first thing to be guaranteed is internal validity. Generally speaking, external validity can not be specially considered - the reason is very simple: as long as there is any difference between the new research situation and the original research situation, external validity may not be guaranteed, and there are too many factors causing the differences between situations, so external validity is difficult to guarantee.

The means to ensure or improve internal validity is control, including experimental control and statistical control. Experiment control mainly refers to doing experiments in the laboratory. Some people may think this is funny. Where can we do experiments if we don't do them in the laboratory? Psychological experiments can be done outside the laboratory. Statistical control, mainly refers to the analysis of covariance. The control of psychological research is actually a way to reduce external validity.

The way to ensure or improve external validity is sampling. Since we don't need to pay special attention to external validity, we don't need to pay special attention to sampling. If someone asks the question that the research sample is not representative, then the researcher can ask the question: how can you prove that the sample is not representative?

If you look at the English literature, you will find that the researchers only describe the subjects, and they don't care about the representativeness of the samples at all. Why? If researchers seem to care, there may be a problem: discrimination! For example, researchers use the research results of Guangxi college students as samples to express the situation of college students, and everyone's intuitive feeling is not good. However, what evidence do you have from scientific research that shows that there are qualitative differences between Guangxi college students and China's college students as a whole?

In psychological research, some differences may be that mathematically, but there are ethical problems. In particular, we should not casually raise the issue of psychological differences in human natural attributes, such as gender, height, weight, body shape, nationality, race, region, etc. Therefore, we can not simply judge that the corresponding research is not representative because of the region of the subjects.

Theoretically, the sample of psychological research should be probability sample; practically, the sample of psychological research is convenience sample. Don't use "random sampling" in your paper! Because researchers can hardly do it, that is, it will not be, and can not be, random sampling. Similarly, in the part of "the shortage of this study and the direction of future research", there is no need to write "this study only investigated the XXX, and we need to further expand the scope of sampling in the future".

I think that the relevant experts should be informed of the situation, otherwise, when they are entangled in this problem in the paper review and / or defense, the author will be in a dilemma - how can the author dare to compete with the experts in such a situation?

Statement: the preciseness of this paper is far lower than that of mathematics. The viewpoints mentioned may inspire the research in the fields of education, sociology and medicine, and may not be applicable to the fields of industry and commerce. In other words, the external validity of this paper is probably low.