期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2018
卷号：96
期号：24
出版社：Journal of Theoretical and Applied
摘要：Small area estimation is a method that arises as a result of the inability of direct estimator to provide an estimator with high precision when the sample size is inadequate. One of small area estimation methods which is widely used to estimate the poverty indicators is a hierarchical Bayes (HB). Yet, the used of HB method needs an intensive computation and requires a full census of covariates or auxiliary variables while in the fact this condition is difficult to reach. This paper proposed a method as the solution of this matter by using the fast hierarchical Bayes approach (FHB) to estimate the poverty indicators. Nevertheless the other problem arises when some of the areas have no sample unit. Therefore, this paper adding the use of cluster information to estimate the area effect. The simulation study is done in two conditions : (1) all the areas have sample units, (2) some areas have no sample units. The evaluation is based on Absolute Relative Bias (ARB) value, Relative Root Mean Square Error (RRMSE) value, and the computational time. Both of the simulation studies showing that the ARB and RRMSE of FHB and HB method is not significantly different. In terms of the computational time, FHB method consumes very less time than the HB method. Thus, it can be said that the FHB method is more effective to use than the HB method especially when the population size is very large. The application study is done to estimate the FGT poverty indicators of the sub-districs in Bogor Regency. The results show that the FHB method can provides better estimator of poverty indicators than the direct estimator, especially in the non-sampled area by adding the cluster information. It can be seen by the posterior variance of FHB estimator which provided very small variances.
关键词：Cluster Information; Non-sampled Area; Fast Hierarchical Bayes;Small Area Estimation; Poverty Indicators
