Also known as classification or type sampling, it is to divide the overall units (or individuals) into different levels or types according to the factors or indicators related to the research content, that is, stratification, and then extract a certain number of individuals from each level in proportion or not, and then form samples by simple random sampling or mechanical sampling. The most commonly used is proportional sampling. The principle of stratified sampling is that the difference within each layer should be as small as possible, while the difference between layers should be large.

For example, a school conducted a sample survey on junior high school students' reading extracurricular books. There are 485 students in the school, including Grade One 180, Grade Two 160 and Grade Three 145. If 100 students are selected from the whole school for investigation, then different grades can be regarded as different levels and divided according to the number of students in each grade. The proportion of students in the three grades in the total number of students in the school is 37% respectively. 33%, 30%, then the number of students in each grade is 37 (that is, 100*37%), 33, 30, and the number of students in each grade can be determined by simple random sampling or mechanical sampling.

For example, there are 500 employees in a company, under 35 years old 125, 35-49 years old 280, and over 50 years old 95. To understand an index related to the physical condition of employees in this unit, we should take a sample with a capacity of 100. Because the age of employees is related to this index, stratified sampling method is adopted for sampling. Because the ratio of sample size to the total number of people is 1: 5, the number of samples of different ages is125/5,280/5,95/5, that is, 25,56, 19.

Generally speaking, when sampling, the population is divided into disjoint layers, and then a certain number of individuals are independently extracted from each layer according to a certain proportion, and the individuals taken from each layer are combined as samples. This sampling method is stratified sampling.

In statistics. It is to divide the whole group into several sub-groups (layers) according to certain characteristics, and then conduct simple random sampling from each layer to form a sample. Multi-layer sampling method refers to the stratified sampling method in which the number of stratified samples in the survey matrix is two or more times. This sampling method is to stratify the survey matrix, then stratify the survey sub-matrix, and finally sample by simple random sampling.