Estimating Poverty Indicator with Small Area Estimation in Simulation Study of Different Population and Sample Size

Abstract

The estimation of poverty indicators of the sub-district or village level can be calculated by small area estimation using direct estimation, empirical Bayes and fast empirical Bayes method. These three methods are evaluated through a simulation study. The usual simulation uses the same population size and sample for each area. This study compares three SAE methods with four population size scenarios with different samples for each area. Based on Bias and MSE values, direct predictions are well used in small populations. The EB method is capable of generating estimation with small bias and MSEs for all scenarios but take longer computation time. While the FEB method produces estimations with bias and MSE are small in large population conditions with faster computational time