Clustering Information of Non-Sampled Area in Small Area Estimation of Poverty Indicators

Abstract

Empirical Bayes (EB) is one of indirect estimates methods which used to estimate parameters in small area. Molina and Rao has been used this method for estimates nonlinear small area parameter based on a nested error model. Problems occur when this method is used to estimate parameter of non-sampled area which is solely based on synthetic model which ignore the area effects. This paper proposed an approach to clustering area effects of auxiliary variable by assuming that there are similarities among particular area. A simulation study was presented to demonstrate the proposed approach. All estimations were evaluated based on the relative bias and relative root mean squares error. The result of simulation showed that proposed approach can improve the ability of model to estimate non-sampled area. The proposed model was applied to estimate poverty indicators at sub-districts level in regency and city of Bogor, West Java, Indonesia. The result of case study, relative root mean squares error prediction of empirical Bayes with information cluster is smaller than synthetic model.