ROBUST SPATIAL REGRESSION MODEL ON ORIGINAL LOCAL GOVERNMENT REVENUE IN JAVA 2017

Winda Chairani Mastuti (Department of Statistics, IPB University, Indonesia)
Anik Djuraidah (Department of Statistics, IPB University, Indonesia)
Erfiani Erfiani (Department of Statistics, IPB University, Indonesia)

Indonesian Journal of Statistics and Its Applications
Vol 4 No 1 (2020)

Spatial regression measures the relationship between response and explanatory variables in the regression model considering spatial effects. Detecting and accommodating outliers is an important step in the regression analysis. Several methods can detect outliers in spatial regression. One of these methods is generating a score test statistics to identify outliers in the spatial autoregressive (SAR) model. This research applies a robust spatial autoregressive (RSAR) model with S- estimator to the Original Local Government Revenue (OLGR) data. The RSAR model with the 4-nearest neighbor weighting matrix is the best model produced in this study. The coefficient of the RSAR model gives a more relevant result. Median absolute deviation (MdAD) and median absolute percentage error (MdAPE) values ​​in the RSAR model with 4-nearest neighbor give smaller results than the SAR model.