人口普查净误差估计中的双系统估计量研究

评估普查计数的完整性已经成为五年或十年一次的人口普查不可分割的一部分。评估通常采取质量评估调查的方式。该调查建立在双系统估计量基础上。考虑到普查登记质量及人口移动因素，这三份人口登记名单可提供7个总体指标。由于质量评估调查采取抽样的方式进行，所以这些总体指标要用样本来构造其估计量。基于复杂抽样方法形成的双系统估计量没有现成的方差公式计算其方差。通常使用分层刀切方差估计量近似计算双系统估计量的方差。这需要计算第一步所有样本抽样单位的复制权数。本文将对双系统估计量构造的各个环节进行理论与实践相结合的阐述，深入解析其中深层次的理论问题，为基础理论研究做出贡献，另外也将探讨基于双系统估计量的合成估计量在区域人口数目估计中的应用问题。

Research on Dual System Estimator for Census Net Error

The evaluation has usually taken on form of quality assessment survey. The survey is based on dual system estimator. Considering census registration quality and mover, three registration lists can provide seven population indicators. Because the quality assessment survey is conducted by sampling, these population indicators must use sample to construct their estimators. There is no existing variance formula to figure variance of dual system estimator, based on complicated sampling method. Variance of dual system estimator can be approximately calculated by means of jackknife variance estimator. This needs to figure all first sampling unit replicate weight. This paper carries out discussion about theory and practice questions for every component of dual system estimator, analyzes their theoretical questions deeply, makes a contribution to its basic theory, and discusses application of synthetic estimator based on dual system estimator to estimate area population.