Definition of sampling survey

It refers to extracting some individuals from the population of the research object as survey samples and inferring the digital characteristics of the population.

Characteristics of sampling survey

Good economy, strong effectiveness, wide adaptability and high accuracy.

Sampling survey is a statistical survey method to infer the total number of overall signs according to some actual survey results, which belongs to the category of incomplete survey. According to the scientific principle and calculation, it extracts some sample units from the population of things composed of several units for investigation and observation, and uses the obtained survey mark data to represent the population and infer the population.

Like other surveys, sampling surveys will also encounter errors and deviations in the survey. There are usually two kinds of errors in sampling survey: one is working error (also called registration error or survey error), and the other is representative error (also called sampling error). However, sampling survey can control the representative error within the allowable range through a series of scientific methods such as sampling design and calculation. In addition, due to the small number of investigation units and strong representativeness, fewer investigators are needed, and the work error is smaller than that of a comprehensive investigation. Especially in the case of a large number of survey units, the accuracy of sampling survey results is generally higher than that of comprehensive survey. Therefore, the results of the sampling survey are very reliable.

Sampling survey data can be used to represent and calculate the population, mainly because the sampling survey itself has the characteristics that other non-comprehensive surveys do not have, mainly:

(1) The survey sample is randomly selected, and the probability of each unit being selected in the population is equal. Therefore, it can ensure the uniform distribution of the selected units in the whole population, and there will be no bias error, which is very representative.

(2) Take all sample units as a "delegation" and use the whole "delegation" to represent the whole. Instead of using randomly selected individual units to represent the whole.

(3) The number of selected survey samples is determined by scientific calculation according to the requirements of survey error, and there is a reliable guarantee on the number of survey samples.

(4) The error of sampling survey can be calculated according to the overall difference between the number of samples and the units before the survey, and controlled within the allowable range, so the accuracy of the survey results is high.

Based on the above characteristics, sampling survey is recognized as the most perfect and scientific survey method used to calculate and represent the whole population among the non-comprehensive survey methods.

Steps of sampling survey

(1) Define population

(2) making a sampling frame

(3) Divide the whole population

(4) Determine the sample size

(5) Determine the reliability and validity of the survey.

(6) determine the sampling method

(7) Conduct sampling survey and guess the whole.

Sampling survey classification

1. Simple random sampling

This is the simplest one-step sampling method, which selects the sampling unit from the population, and every possible sample extracted from the population has the same probability of being extracted. When sampling, the sampling units in the sampling population are arranged into 1 ~ n codes, and then the random numbers between 1 ~ n are determined by a random number table or a special computer program, and those units that match the random numbers in the population become random sampling samples.

This sampling method is simple and easy for error analysis, but it needs a large sample size and is suitable for small differences between individuals.

2. Systematic sampling method

This method, also known as sequential sampling, randomly selects samples from the population at certain intervals (that is, "every few times"). The advantages of this method are good sample distribution, good theory and easy calculation of the overall estimate.

3. Stratified sampling method

According to some specific characteristics, it divides the population into several homogeneous and non-overlapping layers, and then samples independently from each layer, which is an unequal probability sampling. Stratified sampling uses auxiliary information to stratify, each layer should be homogeneous and the difference between layers should be as large as possible. Such stratified sampling can improve the representativeness of samples, the accuracy of overall estimation and the efficiency of sampling scheme, and the operation and management of sampling are more convenient. But the sampling frame is more complicated, the cost is higher and the error analysis is more complicated. This method is suitable for cases with complex matrix, great differences among individuals and large numbers.

4. Cluster sampling method

Cluster sampling is to group the whole unit first, which can be grouped naturally or as needed. In the traffic survey, we can group according to regional characteristics, randomly select groups as sampling samples, and investigate all units in the sample group. Cluster sampling samples are relatively concentrated, which can reduce the investigation cost. For example, in the survey of residents' travel, this method can be used to group residents according to different living areas, and then randomly select the groups as samples. The advantage of this method is simple organization, but the disadvantage is poor sample representativeness.

5. Multi-stage sampling method

Multi-stage sampling is a kind of unequal probability sampling, which requires two or more consecutive stages to sample. The sampling unit of each stage is hierarchical, and the sampling unit of each stage is different in structure. The samples in multistage sampling are concentrated, which can save time and money. The organization of the survey is very complicated, and the calculation of the overall estimate is also very complicated.