The Poverty Modeling Using Small Area Estimation with Semiparametric P-Spline (A case study: Poverty in Bengkulu Province)

Abstract

The main objective of this research is to model poverty in Bengkulu Province using small area estimation (SAE) with semiparametric penalized spline (P-Spline). Small area estimation is a statistical method that is often used to obtain an accurate information about poverty. When the linearity assumption on the basic SAE model is violated, a nonparametric approach is used as an alternative. One is the semiparametric  penalized spline. The small area  method with semiparametric approach has a more flexible model because it accommodates the relationship between response with linear and nonlinear predictors. In this study, poverty modeling in Bengkulu Province was based on average per capita expenditure through the estimation of SAE model parameters using semiparametric P-Spline to obtain a mixed-effect model regression equation as a poverty model. Based on the analysis result, the poverty model in Bengkulu Province is P-Spline linear model with one knot. This model has a GCV value of 148928361265.95. Poverty mapping in Bengkulu Province based on sample villages indicates the estimation of poverty using SAE model with P-Spline having the same trend with the direct estimator.