基于多层次模型的小域估计方法研究OA北大核心CHSSCDCSSCICSTPCD

The core of small area estimation is how to make a more reliable estimation of sub-population characteristics on the area with a very small sample size or even zero.The sample size in a small area is limited,that is,the information available for estimation is limited.Mining the sample information to the greatest extent and inferring the characteristics of this small area with the help of other area samples is the key to improving the accuracy of small area estimation.The traditional design-based infer-ence effects are limited by sample size,which is not suitable for small area estimation with limited sample size.In this case,a mod-el-based method is needed for estimation.For ratio estimation,this paper describes the hierarchical structure between the finite population and the small area based on the multi-level model.The level-one model and the level-two model are used to describe the heterogeneity and correlation between areas,and the sample units of other areas are used to estimate the specified small area.On this basis,the effects of the sampling mechanism and measurement error are examined.For the proposed model,the paper gives the specific parameter estimation and error estimation methods.The specific effects are verified through simulations and ap-plied to real data sets.