人口普查质量评估的三系统模型

Triple System Model for the Population Census Quality Assessment

迄今为止,所有国家和地区在人口普查后组织的用于评估人口普查质量的事后调查均使用双系统估计量估计实际人口数和人口普查覆盖误差。由于人口普查与事后调查不独立而引起的交互作用偏差使双系统估计量低估或高估人口数。文章依据对数线性模型和最大似然估计构造的三系统模型能显著减少交互作用偏差,同时通过一个实证案例演示了三系统模型的选择和实际人口数估计及其方差估计的过程。实证结果表明,人口普查与行政记录独立但均与事后调查相关的三系统模型最适合于既定的数据描述。

So far all countries and regions in assessment of population census' quality with a post-survey have used the dual system estimator to estimate the actual population and census coverage error.That Census and subsequent investigation are not independent with interaction of bias lead to underestimate or overestimate the population.This paper uses log-linear model and maximum likelihood estimation to build a triple system model that significantly reduces correlation bias,and illustrates the selection of triple system model and the process of estimation between the actual population and its variance through a case study.The empirical results show that the census and administrative record system are conditionally independent,but the triple system model of the post-survey is most suitable for the description of established data.