Sampling survey is a kind of incomplete survey, which is a survey method to select some units from all the survey objects and estimate and infer all the survey objects accordingly. Obviously, although sampling survey is not a comprehensive survey, its purpose is to obtain information reflecting the overall situation, so it can also play the role of a comprehensive survey.
According to the sampling method, sampling survey can be divided into probability sampling and non-probability sampling. Probabilistic sampling is based on the principles of probability theory and mathematical statistics, selecting samples from the investigated population, quantitatively estimating and inferring some characteristics of the population, so that possible errors can be controlled in a probabilistic sense.
Several specific sampling methods
1, simple random sampling
Simple random sampling, also known as simple random sampling, refers to a sampling method that randomly selects n units from the total number of n units as samples, so that the probability of each possible sample being selected is equal.
Simple random sampling can generally be carried out by tossing a coin, rolling a dice, drawing lots, looking up a random number table, etc. In the statistical investigation, because there are many overall units, the first three methods are rarely used, and the latter method is mainly used.
Simple random sampling is divided into repeated sampling and non-repeated sampling according to whether each unit is allowed to re-draw. In sampling surveys, especially in socio-economic sampling surveys, simple random sampling generally refers to non-repeated sampling.
Simple random sampling is the basis of other sampling methods, because it is the easiest to deal with in theory, and it is not difficult to realize when the total number of units n is not too large.
But in practice, if n is quite large, simple random sampling is not good. First, it needs a sampling box containing all N cells; Secondly, the sample units obtained in this sampling are scattered, and the investigation is not easy to implement. Therefore, there are not many people who directly adopt simple random sampling in practice.
2. Stratified sampling
Stratified sampling is also called classified sampling or type sampling. First, N units of population are divided into k non-overlapping and non-repetitive parts, which we call layers. Then, each layer of n 1, n2, ... selects nk samples to form a sampling method of sample size.
There are three main functions of stratification: one is for the convenience of work and the need of research purposes; The second is to improve the sampling accuracy; Thirdly, in order to reduce the unit number of samples to save the investigation cost under the requirement of certain accuracy. Therefore, stratified sampling is one of the most common sampling techniques in application.
According to whether the sampling ratio between layers is the same, stratified sampling can be divided into proportional stratified sampling and unequal stratified sampling.
In fact, stratified sampling is an organic combination of scientific grouping and sampling principles. The former is to divide layers with similar attributes to reduce the variation between tag values. The latter is based on the principle of grinding fiber samples to select samples.
Therefore, stratified sampling is generally more accurate than simple random sampling and equidistant sampling, and can obtain more accurate inference results with fewer survey samples, especially when the total number is large and the internal structure is complex, stratified sampling can often obtain satisfactory results.