基于空间模型的小域估计方法

Small Area Estimation based on Spatial Model

小域估计成为当今抽样调查的热点问题之一,日益受到社会各界的关注。小域估计多采用基于模型的估计方法,其中以线性混合模型最为普遍,这种模型通常假定域随机效应是独立的。但是,在实际各个域之间往往表现出一定的空间相关性,并且这种相关性随着距离的增加而减小,若忽视这种空间效应,估计的精度会大大的降低。本文运用域随机效应为空间相关的空间模型来解决空间数据下的小域估计问题,并用基于这种空间模型的权数的方法得到了目标变量的稳健估计量,很大程度上提高小域估计的精度,是一种比较好的小域估计方法。

Small area estimation is one of hottest issues in sample survey, and it gets more and more attention from society. It applies model-based estimation methods, in which linear mixed model is a common model which often assumes that random effect is independent. But in practice, the random effects between the neighboring areas are correlated and the correlation reduces as distance increases. If this spatial effect is ignored, the precision of estimation will be reduced. In this paper, the random effect is taken as a spatial model to solve this problem and the method based on model weight is used to get robust estimator. The precision of small area estimation is increased to a large degree and it＇s a better small area estimation method.