基于混合地理加权回归模型的小域估计  

Mixed Geographically Weighted Regression Model-Based Small Area Estimation

作　　者：李腾 魏传华[1] 苏宇楠[1] 

机构地区：[1]中央民族大学理学院,北京100081

出　　处：《数学的实践与认识》2017年第22期199-207,共9页Mathematics in Practice and Theory

基　　金：国家自然科学基金(11301565);中央民族大学一流大学一流学科经费资助(10301-0170040601-155)

摘　　要：近年来,基于空间模型的小域估计方法在抽样调查领域得到越来越多的关注.提出了一类混合地理加权单元层次小域模型用以刻画空间非平稳性,其中回归系数一部分为常数,另一部分则是随着地理位置的变化而变化.基于模型给出了域均值的估计,并研究了该估计量的均方误差及其估计.Recently, small area estimation based on spatial models have attracted more and more attention in survey sampling. This paper proposes a unit-level mixed geographically weighted regression small area model to capture spatial nonstationarity by allowing some coefficients of the explanatory variables are constant and others are spatially varying. An empirical best linear unbiased predictor for a small area average has been obtained under the proposed model. Finally, the unconditional mean squared error and its estimation of the proposed estimator are developed.

关 键 词：混合地理加权回归模型 混合效应模型 小域估计 

分 类 号：O212.1[理学—概率论与数理统计]