서베이앤헬스폴리시리서치센터

Kim, S. W., Nam, E. J., and Han Y. S. (2012). “A simple approach to sample allocation for multivariate stratified sampling,” Paper presented at the Joint Statistical Meetings, San Diego, California.

Abstract

In stratified simple random sampling, the values of sample sizes in the respective strata should be chosen to reduce sampling variances. For example, if the cost per unit is the same in all strata, Neyman allocation can be used for the purpose. His allocation for minimizing the variance will be the best for one variable, but it will not in general be best for other variables in a survey with many variables (items). Some compromise needs to be reached in the allocation. Huddleston et al. (1970) proposed an algorithm using nonlinear programming (NLP) for compromise allocation. Their classic approach can lead to infeasible NLP problems. Thus, we present a modification of their approach. However, the approach may not be satisfactory because the solution can be less precise than Neyman allocations or proportional allocations for some individual variables. We suggest an alternative approach using NLP based on a different principle. This approach for minimizing the sampling variances of the variables under study is simple to use and always provides a solution. We illustrate the new approach by using real survey data.   

Key Words: nonlinear programming, compromise, precision, sampling variances

For further information on multivariate sample allocation, click on the title of Nam’s thesis (남은정, 2012) or that of Han’s thesis (한영성, 2013) or search for each at https://riss.kr/index.do

비선형계획법을 이용한 층화임의표본의 다중배분에 관한 연구(남은정, 2012)

다변수를 이용한 층화임의추출에서 표본배분에 관한 연구(한영성, 2013)