Pemodelan Mixed Geographically Weighted Regression-Spatial Autoregressive (MGWR-SAR) pada Kasus HIV di Indonesia

Abstract

In general, spatial regression is used to model one of the spatial effects, namely spatial dependency or heterogeneity. For the effects of spatial dependencies, the models that have been used frequently follow Elhost's taxonomy, with the spatial dependencies being on the response, predictor, or error. Whereas for the effect of spatial heterogeneity generally use geographically weighted regression models (GWR) or if there are global predictors use mixed geographically weighted regression (MGWR). The data used in this study are cases of Human Immunodeficiency Virus (HIV) per 100,000 population as a response variable, and key populations, positive cases in pregnant women, tuberculosis patients, poverty rate, and unemployment rate as predictors. In the data used, there are spatial dependencies and heterogeneity. The MGWR-SAR is a model that can be used if the data has both spatial effects. This study aims to determine the factors influencing HIV cases in districts/cities in Indonesia using a spatial model. The results showed that the combined model of GWR and spatial autoregressive regression (SAR) was the best model. Key population explanatory variables have a global and significant influence. Other explanatory variables that have local influence are positive cases in pregnant women, tuberculosis patients, poverty rates, and unemployment rates.