Small area estimation has received a lot of attention in recent years due to growing demand for reliable small area statistics. Traditional area-specific direct estimates based only on sample survey data in the areas of interest do not provide adequate precision for small areas. On the other hand, design-based or model-based indirect small area estimators borrow strength from sample observations of related areas to increase the effective sample size. Supplementary data for small areas of interest such as recent census counts and administrative records are also considered to increase the precision of small area estimators. The main purpose of this thesis is to provide statistical methodology for in-direct estimations in small areas and applications, design-based as well as model-based.
In Chapter 2 the historical background of design-based small area estimation methods such as traditional direct and indirect estimation methods is reviewed in a division of small area statistics. A few examples of these estimators to illustrate the underlying methods are given.
In Chapter 3 model-based small area estimation methods are introduced under assumption of the two types of small area models such as area level and unit level models. Various model-based inferences including Empirical Best Linear Unbiased Prediction (EBLUP), Empirical Bayes (EB), and Hierarchical Bayes (HB) methods are introduced under consideration of the basic area level model.
In Chapter 4 design-based indirect estimation methods such as synthetic and composite estimations are considered to adjust small area unemployment estimates in the Korean Economically Active Population Survey (EAPS). Design problems of these estimators are discussed under the survey system. The jackknife method is employed as an alternative method that provides a more stable and accurate area-specific measure. Model-based indirect estimation methods are also constructed to adjust small area unemployment estimates that derived direc...