Bayesian Geographically Weighted Regression Model for Poverty Data (Case of 35 Villages in Jember Regency).
Model Regresi Terboboti Geografis Bayes untuk Data Kemiskinan (Kasus 35 Desa atau Kelurahan di Kabupaten Jember)
Abstract
Bayesian Geographically Weighted Regression (BGWR) is locally linear regression method to solve some difficulties that arise in Geographically Weighted Regression (GWR) model, such as outliers or non-constant variance. The Bayesian approach solves the problems by producing estimates that are robust against aberrant observations. The aberrant observations are automatically detected and downweighted to mitigate their influence on the estimates. In this research, the weighting used for BGWR model is Gaussian and bi-square kernel function. The results showed that BGWR model is better than GWR model. According to mean square error (MSE) values and coefficient of determinant (R2), Gaussian kernel function is better than bi-square kernel function as BGWR weighting to analyze the data on average expenditure per capita of 35 villages in Jember Regency.