以家庭为单位的抽样调查中总体参数的估计

Parameter Estimation Method for the Data of Family-based Sampling Survey

目的探讨以家庭为单位抽样调查资料的总体参数的估计及其可信区间的计算。方法阐述以家庭为单位抽样调查的特点,以及传统方法处理这类资料的缺陷。构建适当的广义估计方程,采用迭代法估计参数,得到总体参数的正确估计。利用模拟试验进行验证并结合实例介绍该方法的应用。结果传统方法由于没有考虑家庭成员间的相关性,所得到的总体参数的估计存在偏倚,可信区间较考虑层次结构的广义估计方程方法窄。结论广义估计方程可灵活地分析以家庭为抽样单位的资料,且参数估计稳健,可用于评价家庭聚集性、估计参数。

Objective To suggest a parameter estimation method for the data of family-based sampling survey. Methods Monte Carlo Simulation was used to compare a traditional method based on independent binomial data with a method based on Generalized Estimating E- quations（GEEs）. Different incidence rates with different intra-family correlations was considered. An example was illustrated. Results Because of the ignoring of the intra-family correlation, the estimated parameter of tradition method is biased, with lower confidence and more narrow interval than that of GEEs model based when the survey is family-based. Conclusion The parameter estimation method based on GEEs is appropriate and robust when the data is from family-based sampling survey, no matter how much the intra-family correlation. The method can also be used for parameter estimation for cluster sampling data.