The Effects of Extra-Variation to the Estimation of Small-Area Proportions

• 발행기관: 한국자료분석학회
• 저자명: Kwang Mo Jeong Hyun Yung Lee
• 간행물 정보: 『Journal of The Korean Data Analysis Society (JKDAS)』Vol.16 No.6, 2877~2887쪽, 전체 11쪽
• 주제분류: 자연과학 > 통계학
• 파일형태: PDF
• 발행일자: 2014.12.30

영문 초록

In the estimation of small area proportions we frequently encounter small sample sizes and also the heterogeneity between small areas. The binomial-normal logistic mixed model has been used as a model-based method for the estimation of domain proportions. The clustered binomial counts from domains have extra-variation due to the variability between domains. We may consider other kinds of overdispersion models such as the beta-binomial logistic model and the zero-inflated binomial model. We explain the suggested models through an artificial example and also perform a Monte Carlo study for the three populations that are formed under the models mentioned before, that is, the beta-binomial, the zero-inflated binomial and the binomial-normal models. The best linear unbiased predictors of proportions and their mean squared errors would be compared according to the simulated three kinds of populations.

목차

1. Introduction

2. BLUP Type Estimator of Small-Area Proportions

3. A Monte Carlo Study

4. Summary and Comments

References

키워드

small-area estimation extra-variation beta-binomial model zero-inflated binomial model binomial-normal logit mixed model