If the probability of each individual being drawn in the population is equal (that is, the randomness of sampling), and the composition of the population remains unchanged after an individual is drawn by sampling (that is, the independence of sampling), this sampling method is called simple random sampling.
Simple random sampling generally adopts the following three methods:
(1) draw lots. Every individual in the group is numbered and made into a label. After fully mixing, a part is randomly selected, and the individuals corresponding to this part form a sample.
(2) Look-up table method. Check the random number table to determine the number of individuals extracted from the population, and the corresponding individuals will enter the sample. The random number table can start from any number in any area and then proceed in all directions in turn.
(3) Computer-generated numbering method. A random number program is compiled by an electronic computer, which is used as the number of individuals drawn into the sample from the population.
2. Systematic sampling (equidistant sampling)
Systematic sampling method is actually mechanical sampling of equal interval method. It numbers all the individuals in the population in a certain order, and then samples at fixed intervals. The size of the interval depends on the ratio of the required sample size to the number of individuals in the population, and the starting number must be determined randomly. There are three kinds of equidistant sampling: linear equidistant sampling, symmetrical equidistant sampling and cyclic equidistant sampling. Compared with simple random sampling, this method is convenient, easy to learn and easy to operate. When the population is arranged in a certain order, once the first sample is determined, the other samples are also determined. However, if there are periodic parts in the list arrangement, this sampling method will cause deviation. Therefore, after the equidistant sampling interval is determined, when choosing the starting point, we should try to avoid the points in the whole population that may have periods according to the information we have.
3. Stratified sampling (type sampling) Stratified sampling is to divide people into different types or levels according to certain signs, and then randomly select several units from different types to form samples.
The samples taken by stratified sampling at all levels can also be regarded as the apportionment of the total number of samples at all levels, and there are three methods:
(1) proportional sampling, that is, the number of samples in each layer accounts for the same proportion of the total number of units in each layer.
(2) Distribute samples according to the discrete situation of each layer. If a layer has a large dispersion, it will distribute more samples.
(3) Optimal allocation, considering not only the number of units in each layer, but also the dispersion of each layer.
4. Nested sampling method
Cluster sampling is to divide each unit into several groups (groups), and then randomly select some groups as a group unit to investigate all units in the selected group.