基于组合式三系统估计量的人口普查净误差估计

Estimation of Net Error of Census Based on Combined Triple-system Estimator

研究组合式三系统估计量,旨在为政府统计部门设计人口普查净误差估计方案提供重要参考,提高净误差估计精度。研究对象是人口普查标准时点上的人口总体。利用对这个总体登记的普查人口名单、事后计数调查人口名单和行政记录人口名单构造的组合式三系统估计量,解决当今人口普查净误差估计前沿方法的普通三系统估计量所存在的问题。这些问题包括未考虑到不同信息的重要性,以及在使用之前还要进行复杂的检验或选择。为实现这一目标,采取数理模型方法研究组合式三系统估计量及其与之相关的抽样及估计问题。研究结果表明:组合式三系统估计量须在等概率人口层建立,分层刀切法适合于近似计算其抽样方差;事后计数调查人口的构成方法影响组合式三系统估计量的估计精度。采用A构成法建立的组合式三系统估计量在抽样估计精度方面优于使用B构成法构造的组合式三系统估计量;在计算组合式三系统估计量估计的总体实际人口数的抽样方差时,不能忽视协方差的存在,否则低估或高估总体的抽样方差。研究创新是提出抽样登记的、人口移动的组合式三系统估计量。研究价值在于,将目前的普通三系统估计量拓展到组合式三系统估计量,丰富了三系统估计量理论,拓宽了人们研究三系统估计量的视野。建议政府统计部门开展组合式三系统估计量研究,用于未来人口普查净误差估计。

The purpose of studying combined triple system estimator is to provide an important reference for the government statistical department to design population census net error estimation scheme,and to improve the net error estimation precision.The research object is the population at the standard time point of the census.What to study is how to use the combined triple system estimator constructed from census list,post-enumeration survey population list,and administrative record population list,which are used to register the population,to address questions of the common triple system estimator that is the leading edge method for estimating census net error.These problems include failure to take into account the importance of different pieces of information,and complex tests or selections before it be used.To achieve this goal,the method of mathematical model is used to study the combined triple system estimator and its related sampling and estimation.The research results show that the combined triple system estimator must be established in equal probability population strata,and the stratified Jack-knife method is suitable for approximate calculation of its sampling variance;the composition method of the post-enumeration survey affects the estimation precision of the combined triple system estimators.The combined triple system estimator established by the A-composition method is better to the combined triple system estimator constructed by B-composition method in terms of sampling estimation precision;when calculating the sampling variance of total number of people of the population estimated by the combined triple system estimators,co-variance cannot be ignored,otherwise the sampling variance is underestimated or overestimated.The research innovation is to propose a combined triple system estimator with sample enumeration and population movement.The research value lies in extending the current common triple system estimator to the combined triple system estimator,enriching the theory of triple system estimator,and broadening people’s field of study on triple system estimators.It is recommended that government statistical departments study combined triple system estimator for future census net error estimation.