층화다단추출설계의 효율성 평가

Abstract

A stratified multistage sample design is often considered for survey researches. In Korea, many household surveys are conducted by adopting stratified two-stage sample design utilizing so-called “enumeration district (ED for short)” as the primary sampling unit. However, some other alternative area units such as 'Output Area', 'apartment complexes', 'dong-eup-myeon' and so on may be suitable candidates for the primary sampling unit due to many reasons. In this thesis, we study methods to evaluate the efficiency of stratified two-stage sample design and apply them when the “Output Areas“ of the Statistics Korea are used instead as area units in sample selection at the first or later stages through a variance decomposition. Use of Output Areas as clusters leads to attain intra-cluster homogeneity within them to some extent but their clustering effects due to both within-unit homogeneity and size-variation among the Output Areas may be reduced by applying the probability proportionate to size sampling. However, we found out that stratification tends to enhance the precision of the total estimator by excluding between stratum-variations. Our study result shows that the Output Areas may be a plausible choice for designing the stratified multistage sampling in Korean household surveys.