How to estimate the standard deviation and allowable error of a disease sample?
(2) Systematic sampling: it is a method of equidistant sampling in systematic order, which is simpler than simple random sampling. For example, if you select1100 people from 100 people, you can randomly select100 people from the random number table. Suppose the number is 38, 10 000 people. Number from 1 to10 000 people. The distribution of samples in the population is relatively uniform, with good representativeness and small error. Systematic sampling is periodic. If the characteristics of the whole population are also synchronous and periodic, the sampling results will be biased, so in this case, individuals should be extracted with several random numbers or frequent random numbers.
(3) Stratified sampling: When the characteristics of individuals in the crowd are quite different, or you want to know a certain crowd, but the proportion of this crowd to the total population is small, stratified sampling is appropriate. The principle of layering is to try to distinguish between layers, not within each layer. When sampling, each layer must be drawn.
(4) Cluster sampling: This kind of sampling is a random sampling method based on cluster or cluster. For example, in KABP survey, several neighborhood committees in a city are randomly sampled, which is called single-level cluster sampling. If multiple neighborhood committees are selected in a city, and each neighborhood Committee does not fully investigate, but selects several groups of residents to investigate, it is called two-level cluster sampling. Obviously, there can be multi-level cluster sampling. The principle of cluster sampling is that the less differences between groups, the better, and the differences within groups are consistent with the whole. The advantage of cluster sampling is that it is easy to organize, but the disadvantage is that the error is large, so the sample content is larger than that of simple random sampling. In two-stage sampling, it should be noted that the sampling units in the first stage should be basically the same size (that is, the number of sampling units is basically the same, belonging to an order of magnitude, such as tens of thousands or thousands of people), so as to ensure that the total sampling probability remains unchanged. If it is n-level sampling, the sampling units of the first n- 1 level should be basically equal. When the sizes of sampling units differ greatly, they can be adjusted by decomposition and merger to make them basically equal.