Statistics 简明教程
Statistics - Cluster sampling
在 cluster sampling 中，组内本质上是异质的组，并且是随机选择的。不同于 stratified sampling 中组是同质的且从每个组随机选择了几个元素，在 cluster sampling 中，开发了组内异质性的组，并且组内的所有元素都变成了样本的一部分。 stratified sampling 具有组内同质性和组间异质性，而 cluster sampling 具有组内异质性。
In cluster sampling, groups of elements that ideally speaking, are heterogeneous in nature within group, and are chosen randomly. Unlike stratified sampling where groups are homogeneous and few elements are randomly chosen from each group, in cluster sampling the group with intra group heterogeneity are developed and all the elements within the group become a pan of the sample. Whereas stratified sampling has intra group homogeneity and inter group heterogeneity, cluster sampling has intra group heterogeneity.
Examples
One stage cluster sampling
由来自不同部门的成员组成的一个委员会具有高度异质性。当从这样的委员会中随机选择几个委员会时，这就是 one stage cluster sampling 的情况。
A committee comprising of number of members from different departments has a high degree of heterogeneity. When from number of such committees, few are chosen randomly, and then it is a case of one stage cluster sampling.
Two stage cluster sampling
如果从每个随机选择的集群中使用简单随机抽样或任何其他概率方法随机选择几个元素，那么它就是 two stage cluster sampling 。
If from each cluster which has been randomly chosen, few elements are chosen randomly using simple random sampling or any other probability method then it is a two stage cluster sampling.
Multi-stage cluster sampling
当样本中元素的选择涉及多个阶段的选择时，群集样本可以是多阶段抽样，例如，如果在国家保险产品调查中需要抽取保险公司的样本，则需要在多个阶段开发群集。
A cluster sample can be a multiple stage sampling, when the choice of element in a sample involves selection at multiple stages e.g. if in a national survey on insurance products a sample of insurance companies is to be drawn, then it requires developing clusters at multiple stages.
在第一阶段，群集是根据公私公司形成的。在下一阶段，从先前开发的每个群集中随机选择一组公司。在第三阶段，从收集数据的每个所选公司的办公地点中随机选择。因此，在多阶段抽样中，对初级单位执行概率抽样，然后从每个初级单位中抽取二次抽样单位的样本，然后第三层，直到我们达到样本单位的细分最终阶段。
In the first stage the clusters are formed on the basis of public and private companies. At the next stage a group of companies is chosen randomly from each cluster developed earlier. In the third stage the office location of each chosen company from where data is to be collected is chosen randomly. Thus in multistage sampling, probability sampling of primary units is done, then from each primary unit a sample of secondary sampling units is drawn and then the third levels till we reach the final stage of breakdown for the sample units.