Perbandingan Regressi Data Panel Dan Geographically Weighted Panel Regression Terhadap Persentase Penduduk Miskin di Jawa Timur

Abstract

This study aims to determine the comparison of the GWPR method with panel data regression and identify factors that affect the percentage of poverty in East Java Province using the GWPR method and panel data regression. Panel data regression uses Ordinary Least Square (OLS) in estimate parameters, while in GWPR parameter estimation use Weighted Least Square (WLS). First, carry out a multicollinearity test, then proceed with modeling panel data regression. The spatial weighting on GWPR model was calculated using fixed bisquare and fixed gaussian. The optimum weighting function is fixed bisquare which provides a minimum Cross Validation (CV) value of 0.01169005. The result showed that panel data regression produces a global model, while GWPR produces a local model. The local model in GWPR means that each regency/municipalities has a different model. Result of the comparison is GWPR was more accurate than panel regression with a good model of 0.9985947 and RMSE value of 0.1680463. These factor are namely unemployment rate, mean years of school, life expectancy, access to proper sanitation, non-electric lighting source, adjusted real per capita expenditure