双系统估计量与人口普查内容误差评估

Dual-System Estimator and Census Content Error Evaluation

内容误差的评估对象是人口普查标准时点上所发生的普查项目填写错误及普查多报与漏报。比率内容误差估计量是目前评估内容误差的主要方法。这种方法存在两个缺陷:一是评估的是样本本身的内容误差,从而违背了抽样理论及内容误差评估的目标。二是未包括普查多报与漏报引起的内容误差,从而低估内容误差。针对这一问题,拟提出用"完整测度的内容误差估计量"予以替代的研究目标。该内容误差估计量覆盖了普查多报、漏报及普查项目填写错误引起的全部内容误差,发展了内容误差评估理论。为实现目标,采用基于捕获再捕获模型的双系统估计量研究"完整测度的内容误差估计量"及其分层刀切抽样方差估计量。研究表明:"完整测度的内容误差估计量"将引发内容误差的各种来源置于测度的范围之内,能对普查登记质量做出全面客观科学评估,提高总体内容误差估计精度;为避免异质性偏差,建立"完整测度的内容误差估计量"时须对人口组的人口进行等概率分层,而不能直接在人口组构造。尚未从现有相关文献资料中找到"完整测度的内容误差估计量",因此该估计量是原创的。建议中国和其他一些国家的政府统计部门加强"完整测度的内容误差估计量"的基础理论研究和应用研究,将其应用于未来人口普查内容误差评估工作中。

The evaluation object of content errors are incorrect census item filling,census erroneous enumerations,and census omissions that occurred at the standard time point of the census.The ratio content error estimator is currently a main method of evaluating content error.This method has two drawbacks.One is that the content error of the sample itself is evaluated,which violates the sampling theory and the goal of content error evaluation.Second,the content error caused by census erroneous enumerations and census omissions is not included,thereby underestimating the content error.In response to this problem,one research goal is proposed,which is to replace the ratio content estimator with"complete measurement content error estimator".This content error estimate covers all content errors caused by census erroneous enumerations,census omissions,and census item filling errors,and develops content error evaluation theory.To achieve the goal,a dual-system estimator based on the capture-recapture model is used to study the"complete measurement content error estimator"and its stratified Jack-Knife sampling variance estimator.Theoretical and empirical analysis results show that:"complete measurement content error estimator"puts the various sources of content error within the measurement range,can make a comprehensive and objective scientific assessment of the enumeration quality of the census,and improve the accuracy of the overall content error estimation;To avoid heterogeneity bias,when establishing the"complete measurement content error estimator",the people of the population group must be stratified with equal probability,and cannot be directly constructed in the population group."Complete measurement content error estimator"has not been seen in the existing related literature,so the estimator is original.It is recommended that the government statistics departments of our country and some other countries strengthen the basic theoretical research and application research of the"completely measured content error estimator",and apply it to the future census content error evaluation work.