本文提出以事後分層法為依據的小區域估計值。該估計值同時考量層內同質性假設之合理性以及小區域樣本數不足的困境。我們推導簡單隨機抽樣下該估計值的小樣本性質。在超族群模式下，經由貝氏分析可以證實估計值的合理性。此外，我們探討複雜抽樣下，該估計值的推廣。 A compromise estimator employing poststratification is proposed for small area estimation. The estimator strikes a balance to deal with the assumption of similarity within a poststratum and small number of observations in subareas constructed by poststratification. Small sample properties are derived for the estimator from simple random samples. By assuming that the probability that poststratum size equal to zero is negiligibly small, the estimator can be justified in Bayesian terms based on an inherent superpopulation model. The generalization of poststratification from the simple random sampling to complex design is discussed.
