半参数乘积调整模型的抽样估计方法及应用研究  

The Method and Application Research on Sampling Estimation Based on Semi-parametric Product Adjustment Model

作　　者：陈光慧[1] 曹伟伟 CHEN Guang-hui ,CAO Wei-wei(School of Economics, Jinan University, Guangdong Guangzhou 510632, Chin)

机构地区：[1]暨南大学经济学院

出　　处：《数理统计与管理》2018年第3期449-458,共10页Journal of Applied Statistics and Management

基　　金：国家社会科学基金项目(14CTJ014);国家“万人计划”青年拔尖人才支持项目的阶段性成果

摘　　要：在抽样调查中，经常会遇到研究变量与其相关的辅助变量之间呈现非线性关系的情形，这时应用传统的广义回归估计方法将存在较大的局限和不足。为了解决这一问题，本文首先假定超总体模型的回归函数为一般形式的光滑函数，从而构建出半参数回归模型，利用半参数乘积调整方法对模型进行拟合，结合广义差分估计原理提出一种新型的半参数模型辅助抽样估计量，通过理论证明和数值模拟分析分别对比验证了这一套估计方法与传统方法相比所具备的有效性。在此基础上，对于在我国社会经济统计中如何应用这一套估计方法进行了展望。In sampling survey, there is often a nonlinear relationship between the study variables and their associated variables. Then, generalized regression estimator would perform inefficiently. In order to solve this problem, we firstly assume that the regression function of superpopulation model is a smooth function and propose a semi-parametric model, by which the semi-parametric product adjustment method is applied. Combining with generalized difference estimation, a semi-parametric model-assisted estimator is obtained. The validity of this set of estimation methods compared with the traditional methods is verified by theoretical and numerical simulation. On that basis, this paper predicts how to apply this set of estimation methods in the social and economic statistics of China.

关 键 词：抽样估计 模型辅助估计 超总体回归模型 广义差分估计 半参数乘积调整 

分 类 号：C811[社会学—统计学] O212[理学—概率论与数理统计]