Small Area Estimation Method And Empirical Research Of Sampling Survey

Small domain estimation is an important research direction in the field of sampling survey.Many important issues in national economy and people’s livelihood need to be studied by small domain estimation methods,such as unemployment rate,crime rate,and disability rate.Small domain estimation originates from abroad.Its theoretical and practical application research is relatively more abroad,less domestic,and lags behind the international advanced level.Therefore,this paper studies the theory and application of small domain estimation.In the theoretical research of small domain estimation,the basic methods of small domain estimation are systematically summarized.In order to solve the problem that the small-domain estimated value cannot be obtained by direct estimation due to the insufficient sample size of the small-domain,it can be improved from the perspective of expanding the sample size and improving the estimation technology.The basic methods of small domain estimation are summarized mainly from the perspective of improving the estimation technology,and the small domain estimation methods are divided into two categories: direct estimation methods and indirect estimation methods.Indirect estimation methods are divided into implicit model estimation methods and explicit model estimation methods.This article selects the prevalence of diabetes among the elderly in various provinces in China as an indicator of the health level of regional residents,and selects the per capita GDP of each province as an indicator of the level of regional economic development.Using CHARLS 2015 data and China Statistical Yearbook data,the small field estimation method of the unit level Logistic domain random effect model and the domain level Bayesian hierarchical model was used to estimate the prevalence of diabetes in the middle-aged and elderly population in various provinces in China.The relationship between the prevalence and regional economic development is analyzed from the perspective of the correlation between the prevalence and per capita GDP andthe analysis of spatial distribution.The study found that the Logistic domain random effect model at the unit level,because it uses detailed individual level information,has a strong ability to explain the model,and has obtained a relatively reliable prevalence estimation result;the Hierarchical Bayesian model at domain level has the characteristics of robustness,it has a certain model averaging effect,can take into account the regional differences and homogeneity,and it is also more reliable than simple small domain estimation,and its estimation effect is slightly better than that Logit model at domain area.The research results shows that the model-based small domain estimation method can overcome the large deviation of direct estimation results due to a small sample size to a certain extent,reflecting the advantages and strong applicability of the small domain estimation method.In addition,studies have shown that the prevalence has a slight upward trend as a whole as the level of regional economic development increases.We are warned that in areas with high levels of economic development,the middle-aged and elderly in China must pay special attention to the prevention and control of chronic diseases such as diabetes,strengthen physical exercise,and reasonable diet.