Research On Small Area Estimation Methodology Based On Complex Sampling And Its Application In China

Small area estimation methodology has been applied to agricultural statistics,social statistics and so on. Research on small area estimation at home and abroad focuses mainly on one-stage and one-time sampling survey, but our government mainly adopts multistage and successive sampling. Therefore, it is of great importance to study small area estimation methodology based on multistage and successive sampling. Sampling weights are introduced into small area estimators.Basic models and estimation methods are applied to multistage and successive sampling survey. Area level model, ultimate sampling unit-level model and primary sampling unit-level model are formulated according to the level of auxiliary information. Small area estimation models are formulated according to rotation scheme, which include primary sampling unit-level rotating panel survey and ultimate sampling unit-level rotating panel survey. A sampling scheme and a set of estimation methods are formulated on the sampling survey of planted area of certain crops of counties. The results of data simulation show that estimators in this paper have bigger variances and smaller biases than PEBLUP. When estimators of variances of random domain effects are nonpositive, estimators in this paper are still available. In successive sampling, the estimation accuracy of small area estimators can be effectively improved by taking small area estimators of previous occasion as auxiliary information. The methodology in this paper can be applied to crops sampling survey in China.