Geographically weighted regression with fixed kernel tricube on life expectancy in East Java

Violation of the homoscedasticity assumption will cause biased and inefficient parameter estimation. Geographically Weighted Regression (GWR) is a spatial regression analysis that will produce local parameter estimates. Data on life expectancy for East Java Province in 2021 shows that there is a violation of the homoscedasticity assumption. The aim of the study is to analyze Geographically Weighted Regression with fixed kernel tricube for life expectancy in East Java Province in 2021. This study use a non-reactive research design using secondary data. The results show that the GWR model is different from the OLS model and produces four regional groups which are divided based on significant predictor variables. Three significant predictor variables are the percentage of households that have access to proper sanitation (X1), the coverage of exclusive breastfeeding for infants aged less than 6 months (X2), and the percentage of monthly expenditure per capita for food (X5). Based on this study, group mapping is obtained showing that areas in the same group tend to be close together and parameter estimates that are local to the observation locations.