Metode Prediksi Tak-bias Linear Terbaik dan Bayes Berhirarki untuk Pendugaan Area Kecil Berdasarkan Model State Space [DISERTASI]

Abstract

There have been two main topics developed by statisticians in a survey, i.e. sampling techniques and estimation methods. The current issues are estimation methods related to estimation of a particular domain having small size of samples or, in more extreme cases, there is no sample available for direct estimation. Sample survey data provide effective reliable estimators of totals and means for large area and domains. But it is recognized that the usual direct survey estimator for a parameter of small area, have unacceptably large standard errors, due to the circumstance of small sample size in the area. The most commonly used models for this case, usually in small area estimation, are based on generalized linear mixed models. Some time happened that some surveys are carried out periodically so that the estimation could be improved by incorporating both the area and time random effects. In this dissertation we propose a state space model which accounts for the two random effects and is based on two equations, namely transition equation and measurement equation. Based on an evaluation criterion, the proposed hierarchical Bayes estimator turns out to be superior to both estimated best linear unbiased prediction (BLUP) and the direct survey estimator. The posterior variances which measure accuracy of the hierarchical Bayes estimates are always smaller than the corresponding variances of the BLUP and the direct survey estimates