Research On The Application Of Three-System Estimator In The Net Error Estimation Of The Census

After each census registration work,its quality must be assessed.One of the tasks of census quality assessment is to estimate the net error of the census.It is defined as the difference between the total actual population of the census target and the registered population of the census.Since the number of population registered in the census is known,the key to the net error estimation is to obtain an estimate of the total actual population.The research on the academic history of census quality evaluation has always been around this question.The current practice of our country and many other countries is to estimate the actual population using a dual-system estimator constructed from the census population list of the same sample census district and the quality assessment survey population list.Studies have shown that the dual-system estimator contains interaction bias,which underestimates or overestimates the actual population.How to solve the problem of interaction bias has become a frontier issue in the field of net error estimation in the international census.In response to the above-mentioned problems,the goal of this article is to add the administrative record population list on the basis of our country’s current practice,and to construct a three-system estimator to replace the current dual-system estimator by juxtaposing the two previous population lists.The two-system estimator is transplanted from the capture-recapture model originally constructed by biologists to estimate the overall size of wild animals,while the three-system estimator is transplanted from the three-capture model.In the census quality assessment,the population register list of the census is regarded as the first capture,and the population register list of the quality assessment survey is regarded as the second capture,to establish a dual-system estimator for estimating the target population;If the population registration list of the population administrative record system is further regarded as the third capture,the population registration lists of these three systems can be used to establish a three-system estimator for estimating the target population.To achieve the above goals,the following contents are studied by using the methods of literature interpretation,sampling estimation,investigation and simulation research: sample design and data collection process,sample size calculation and distribution,as well as the acquisition and comparison of the same sample community census population list,quality assessment survey population list and administrative record population list,which is a three-system estimation and other estimates are established to provide basic data;the triple capture model for estimating the net error of the census,the relationship between the triple capture model and the three-system estimator,the choice of the three-system estimator,the construction of the three-system estimator in the equal probability population layer and the overall structure and the estimator of the sampling variance,estimate the composite estimator of the net error of the small area census and its variance,the dual-system estimator and the single-system estimator based on three population lists.The study found that the three-system estimator must be established in the equal probability population level;the three-system estimator can be constructed according to the statistical relationship between the three lists;the three-system estimator is a complex estimator,and the sampling variance is approximately calculated by stratified jackknife method;maximum likelihood estimation,log-linear model,proportional iterative fitting and incomplete three-dimensional contingency table are important tools for constructing three-system estimator.Research innovation: proposing a post-stratification plan suitable for China;improve the technical details based on the net error estimation of the three-system estimator;discuss the comparability of the three-system estimator,the two-system estimator,and the single-system estimator.The research value is reflected in two aspects.First,the current dual-system estimator has a flaw that cannot be ignored,that is,the interaction bias caused by the inability of the census and the quality assessment survey to be independent in reality makes it systematically underestimate or overestimate the actual population,but introduces the three-system estimator of the population administrative record system can be constructed either in the case of independent or non-independent three systems.Therefore,the estimated population and the net error of the census are more consistent with the actual situation,and the accuracy is higher.This will help the Chinese government and relevant scientific research institutions to correctly use the census data and make an objective and scientific evaluation of the quality of the census.It also helps the development and utilization of administrative records in our country,and create conditions for establishing a census system with administrative records in the future;secondly,it analyzes the feasibility of applying the three-system estimator in China’s census from two aspects: theoretical research and practical application.Attempt to introduce population administrative records on the basis of the dual-system estimator and construct a three-system estimator.Use the hierarchical Jackknife method to estimate the variance of the three-system estimator.The simulation analysis demonstrated the specific steps of the three-system estimator,and provided specific algorithms for the application of the three-system estimator in our country.