Overall Estimate Of The Population Record Linkage Hierarchical Bayesian Model

With the development of the society and the advancement of technology, our country after the founding of the People’s Republic of China has successively carried out six censuses. Through the census, we can collect population data and obtain the national total population. Then we can work out relevant education policies, population employment, national policy, social security and welfare policy, the elderly population policy and so on. While the formulation of these policies will depend on the real total population, although every country has carried on the careful plans and arrangements before the census, but the census is impossible to count everyone, inevitably there are omissions, also clerks at the census would inevitably have error, so the estimate of the real total population is worthy of in-depth discussion and research.To get the correct population, it is necessary to select the correct estimation method. At present, most countries in the world use the post enumeration survey of the census to estimate the real population. This paper systematically combed the methods used to estimate the real population in recent years. Then we introduce a hierarchical Bayesian approach to record linkage for matching statistical records observed on different occasions from Andrea Tancredi and Brunero Liseo (2011). And we mainly study the matching process between census data and post enumeration survey data.First of all, this paper describes in detail the common approach of estimating real population-dual system estimator. In chapter3, we describes record linkage model used in this paper and introduces the background of record linkage and record linkage model-Fellegi-Sunter model. What’s more, we summarize the shortage of the model. Then we describe our new model-a hierarchical Bayesian model to record linkage. We show this model can be profitably adopted both in record linkage problems and in capture-recapture setups. In parameter estimation stage, we describe the Gibbs sampling update steps in the process of model parameter estimation in detail. When using a loss function that justifies the false match rate (FMR) as a measure of performance of record linkage techniques, we describe several specific loss function. Finally, we illustrate our approach through a simulation data and summarize the deficiency of this article.