1. Pure random sampling: all the capacities of the population are not sorted and queued, and the sample capacities are sampled one by one according to the principle of randomness.
Common sampling methods of pure random sampling
1) lottery method: all the total capacities are numbered and compiled into corresponding figures, and then the figures are fully mixed and extracted one by one until the predetermined sample capacity is extracted.
Disadvantages: when the overall capacity is large, the workload of numbering labels is very large and it is difficult to mix them evenly.
2) Random number method: use any convenient method, such as alphabetical order or ID number, to number the total capacity, use a random number table to randomly select the number of n (sample size) from the total capacity, and skip those numbers that are not in the number.
Second, equidistant sampling: first, line up the whole units according to a related symbol (or irrelevant symbol), and then sample the sample units at equal distances or intervals. According to the number of sample units (n) and the total number of units (n), the distance and interval between sample units can be calculated, that is, K=N/n, and then the required sample units can be extracted in turn according to this interval.
An example of equidistant sampling
There are 5000 employees in an enterprise, and now 100 people are randomly selected for family income survey.
Extraction method: arrange the crowd according to the strokes of names that are not directly related to the research purpose, divide the crowd into K=5000/ 100=50 equal intervals, and randomly extract a person from 1 to the 50th one, such as 10, and then extract the 60th one from the following intervals,160.
Advantages of equidistant sampling: (1) can ensure that the sampling units are evenly distributed in the population; (2) Simplify the sampling process.
Pay attention to equidistant sampling: avoid the coincidence of sampling interval or sampling distance with the rhythm or cycle period of the phenomenon itself.
Three. representative sampling
Type sampling: group the whole and all the units in the whole according to a main sign, and then select a certain number of investigation units in each group by pure random sampling or equidistant sampling to form the required samples.
Scope of application: it is mainly applicable to situations where the overall situation is complex, there are great differences among various types or levels, and there are many overall units. Layering reduces the differences between units in layers and enlarges the differences between layers.
(1) proportional sampling type
Distribute the number of sampling units in each group in proportion to the total number of units in each group. That is, the ratio of the number of sample units ni extracted from each type to all units Ni of that type is equal, which is equivalent to the ratio of the total number of sample units n to the total number of units n, that is:
The number of sample units drawn from each type group is:
Sampling ratio Sampling sample size: ni units are selected from each group of Ni units according to the predetermined ratio (n/N) to form a sampling population, and the sample size is:
n= n 1+ n2+ n3+…+ nk=
(2) Types suitable for sampling
When sampling the number of sample units, it is necessary to consider that the number of units contained in each type group is different and the degree of variation () of symbols is different. The type group () with large variation degree should sample more units, and the type group () with small variation degree should sample less units, so that the proportion of variation degree () of each type group to the sum of variation degrees of all types groups is equal, equal to or.
In addition, the number and degree of change of units in various types of groups can be considered together, so that the proportion in the sum of all types of groups is equal to or, that is:
Fourth, cluster sampling.
Taking groups (or groups) as a unit in the whole and population, several groups (or groups) are extracted by pure random method or equidistant sampling method, and then all units in all the extracted groups (or groups) are investigated one by one.
Five, multi-stage sampling
Divide multiple sampling procedures into several stages, and then sample step by step to complete the whole sampling process.
Scope of application: there are many overall units and they are widely distributed. It is very difficult to select samples through one sampling, so multi-stage sampling is adopted at this time.
An example of multi-stage sampling
Sampling survey of agricultural output in China.
The sampling method is as follows: firstly draw counties in the province, then draw townships and villages in the county, then draw plots from the drawn townships and villages, and finally draw units from the drawn plots.