The research object is the population(national and sub-national)at the standard point time of the census.It is examined how to estimate the number of census omissions using a combined-omission estimator constructed from the census population list,the coverage survey population list,and the matched population list of these two survey population lists.The research objective is to implement the proposed combined-omission estimator to replace the single-omission estimator currently recommended by the United Nations Statistics Division for countries.At present,government statistical departments in various countries mainly use the single-omission estimator in population census omission estimation.This omission estimator is established by the omitted population registered on the coverage survey population list but not on the census population list and its sampling weight.The single-omission estimator does not include the double omitted population that is omitted from both survey population lists,so it underestimates the overall population omitted in the census.The combined-omission estimator not only includes the single omitted population but also includes the double omitted population,thus avoiding the defects of the single-omission estimator.The combination method of sampling estimation and mathematical modeling is used to study the single-omission estimator,the double-omission estimator,the combined-omission estimator,and their biases,sampling variance and mean square error estimation.Results show that:The combined-omission estimator consists of a single-omission estimator and a double-omission estimator;The theoretical basis of the double-omission estimator is the capture-recapture model,which must be established in the equal probability population stratum,while the single-omission estimator does not need to be constructed in the equal probability population stratum;To facilitate the establishment of the combined-omission estimator,the single-omission estimator and the double-omission estimator can be established in the same equal probability population stratum;Under stratified two-phase sampling,the component elements of the combined-omission estimator are constructed by the double-expansion estimator;The combined-omission estimator is relatively complex,and its sampling variance is approximately estimated by the stratified jack-knife sampling variance estimator;Considering the correlation between various equal probability population strata,when calculating the sampling variance of the combined-omission estimator for the population,in addition to calculating the sampling variance of each equal probability population stratum,the co-variance between them should also be calculated;The single-omission estimator,the double-omission estimator,and the combined-omission estimator are all biased,and their estimation precision should be compared using mean square error;And the combined-omission estimator is superior to the single-omission estimator in sampling estimation precision and estimating the level of population omission,especially in regards to analyzing the characteristics of a double-omission population and improving future census registration.The innovation lies in the use of raw data to comprehensively demonstrate the detailed calculation process of the single-omission estimator,the double-omission estimator,the combined-omission estimator,and their sampling variance estimators.Statistical properties are discussed,especially using the hypothetical data to simulate bias.Using the case of China,the results contribute a novel method to the scientific formulation population census omission estimation scheme and to improve estimation precision.
万方数据
维普
