Sampling is related to the reliability and effectiveness of research. The basic principle of sampling design should be randomness, that is, when sampling, the probability of each individual being selected in the population is completely equal, so sampling is likely to maintain the same structure as the population, thus ensuring the representativeness of the sample. In addition, random sampling can also use statistical methods to budget and control the range of sampling errors, so that researchers can objectively evaluate the accuracy of research results and determine the sample size according to the required accuracy.
(2) Sampling procedure
(1) Define the crowd: define the crowd according to the purpose of the study and explain the specific connotation of the crowd. The general provisions affect the research value, work content and cost, and the external effectiveness of the results, which is of great significance;
② Determination of sample size: Calculate the number of subjects included in the study according to the statistical principle. The ideal sample size is to control the research cost by minimizing the number of objects under the premise of meeting the requirements of statistical representation;
③ Determining sampling methods and implementing sampling: designing specific sampling methods and implementing them according to sampling principles and conditions;
④ Statistical inference: Explain how this study estimates the population parameters through sample statistics and statistical methods.
(3) Sampling method
Researchers should choose appropriate sampling methods according to the needs of different research purposes and research conditions. Commonly used random sampling methods include:
① Simple random sampling method: according to the principle of randomness, several units are directly extracted from the population as samples to ensure that every object in the population has the same possibility of being extracted, and they are required to be independent. The specific extraction methods are lottery and random number table. The random sampling method conforms to the principle of probability theory in theory, and the error calculation is convenient. It is suitable for studying the situation that the proportion of all kinds of individuals in the population is unknown, or the differences between individuals in the population are small, or the number of samples is large. But its limitations are: when the sample size is small, there may be deviation, which will affect the representativeness of the sample; It is impossible to control a certain characteristic of the research object that is known to directly affect the research results.
(2) Systematic random sampling method: firstly, all the units in the group are arranged and numbered according to a certain symbolic order, then the number of units in the group is divided by the number of units in the sample to obtain the sampling interval, and finally, a unit is randomly selected as the first sample unit in the first sampling interval, and equidistant sampling is carried out according to the sampling distance until the last sample unit is extracted. Systematic random sampling can systematically sample the whole population, so the sample is more accurate. Generally speaking, its sampling error is smaller than that of simple random sampling. However, if the population fluctuates or changes periodically, systematic deviation may occur.
③ Stratified random sampling method: firstly, divide the whole unit into several types according to certain standards, then determine the number of sample units to be extracted from each type according to the ratio of the number of type units to the total number of units, and finally extract samples from each type according to the random principle. Classification should follow the principle of small intra-layer variation and large inter-layer variation, and each unit should belong to a specific category. The number of samples taken from each layer should be the product of the ratio of the layer to the total and the total sample size. The advantage of stratified random sampling method is that it has good representativeness and accuracy of inference, and it is suitable for the research objects with a large number of overall units and large internal differences, and can control the sampling error with a relatively small number of samples. In addition, different sampling methods and proportions can be adopted for each layer according to specific conditions, which makes sampling more flexible. Therefore, it is a very common sampling method. However, the limitation is that the classification is very scientific, and sometimes it is necessary to estimate the standard deviation according to previous data or research, so it is more complicated.
④ Cluster random sampling method: firstly, the whole unit is divided into many groups according to certain standards, and then some groups are taken as samples from these groups according to the random principle. Its advantage is that the sample units are concentrated, which is suitable for some specific research. For example, in the teaching experiment, the study is carried out in class. In large-scale investigation and study, cluster random sampling is easy to organize and can save manpower, material resources and time. Its disadvantage is that the sample distribution is uneven, which is affected by the differences between groups.
⑤ Multi-segment random sampling method: firstly, the research units are divided into several groups as primary units for sampling according to certain standards, then the primary units are divided into several subgroups as secondary units for sampling according to certain standards, and so on, and samples are drawn from units at all levels according to the principle of randomness. In short, it is a sampling method that divides the process of sampling from the population into two or more stages. Multi-stage random sampling method can comprehensively use various sampling methods, which is simple and economical, and is very useful in the case of large research scope, many units and complicated situations. Its disadvantage is that the sampling error is larger than that of simple random sampling.