복잡표본설계를 위한 설계효과모형 및 설계요소 평가

Abstract

The design effect is defined by a ratio of the variance of the estimates of interest under a complex sample design to the variance of a hypothetical simple random sample of the same sample size. It is often used to evaluate the effectiveness of design components of the exiting surveys or to plan new sample designs. Complex sample design usually contains complex design features such as stratification, clustering, and unequal weighting. Kish(1987) presented design effect model in the form of multiplication for each model of the clustering and the unequal weighting effects based on intuition. Gabler et al.(1999) gave justification to Kish’s design effect model based on the one-way random effect model. We study the design effect model by Park(2015), Chen and Rust(2017) that extends of Gabler et al(1999)’s model to include stratification effects as well as Kish’s model. A limited simulations are carried out to assess the design effect model and the design component estimates under stratified multistage sample design.