常用二阶段抽样方法的评价

摘要

Objective To evaluate the bias generated by three different two-stage sampling methods in epidemiological survey. Methods There are three sampling methods in cross-section survey, which are random sampling, sampling in the central location according to prior information, convenience sample. The 95% confidence intervals are estimated according to the sampling data got by three different sampling methods. The coverage rates of the population mean by different sampling method are estimated and compared. Results The coverage rates of the population mean by different sampling methods are given in each case. Conclusion When the ICC(intraclass correlation coefficient) is small,sampling in the central location according to prior information and random sampling have the similar effect. The effect of these two methods is different with the method of convenience sample. So, We can use the method of sampling in the central location according to prior information instead of random sampling when the ICC is small.%目的 评价在流行病学调查研究中,三种常用的二阶段抽样方法对估计总体参数产生的偏倚.方法 在横断面调查的二阶段抽样中,常用的抽样方法有随机抽样、根据先验信息在中心位置抽样、以及方便抽样等.用SAS软件编程对三种抽样方法得到的各个样本资料进行区间估计以及估计95％可信区间包含总体均数的覆盖率,比较不同抽样方式的覆盖率差异.结果 分别给出了在三种抽样方法下得到的区间估计包含总体均数的覆盖率.结论 在ICC(组内相关系数)比较小时,根据先验信息在中心位置抽样与随机抽样具有相似的效果.在这种情况下,可以选择根据先验信息在中心位置抽样或随机抽样的方法.大多数情况下,方便抽样的偏倚最大,这种方法在实际中不建议使用.