서베이앤헬스폴리시리서치센터

Kim, S. W., Heeringa, S. G., Hong, S. J., and Solenberger, P.W. (2008). “Optimized whole sample procedures versus traditional draw-by-draw procedures,” in Proceedings of the Survey Research Methods Section, American Statistical Association, 2825-2828, Denver, Colorado: Joint Statistical Meetings.

Abstract

Kim, Heeringa, and Solenberger (2006) suggested model-based sampling methods for reducing the variance of the Horvitz-Thompson (1952)’s estimator. Their methods are whole sample procedures based on optimization theory under a superpopulation model. With respect to the sample selection probabilities, we theoretically present the differences between those procedures and popular draw-by-draw procedures such as the sampling methods of Mizuno (1952), Brewer (1963), and Murthy (1957), when the sample size is two, which is a common situation in nationwide samples with many strata. We also compare the efficiencies between them for natural populations in the published literature. In addition, we empirically compare the properties of the sample selection probabilities obtained from the procedures through a graphical display.

Key words: Superpopulation Model, Expected Variance, Joint Probabilities

Please check out the file titled JSM2008 Slides. It contains interesting results (pp.33-36) presented through a graphical display.