Geographically Weighted Panel Regression Modelling of Dengue Hemorrhagic Fever Data Using Exponential Kernel Function

Abstract

Geographically Weighted Panel Regression (GWPR) model is a panel regression model applied to spatial data. This research takes the Fixed Effect Model (FEM) panel regression as the global model and GWPR as the local model for dengue hemorrhagic fever (DHF) in East Kalimantan Province data over the years 2018-2020. DHF is a disease that has the potential to become an extraordinary event which is accompanied by death. In comparison to Indonesia, East Kalimantan Province's DHF Incident Rate (IR) was high in 2020. East Kalimantan's IR is 60.6 per 100,000 population, compared to Indonesia's IR of 40.0 per 100,000 population. This research aims to obtain the GWPR model, as well as to acquire factors that affect DHF in East Kalimantan Province over the years 2018-2020 based GWPR model. The parameter of the GWPR model was estimated on each observation location using the Weighted Least Square (WLS) method, which is an Ordinary Least Square (OLS) with the addition of spatial weighting. The spatial weighting on the GWPR model was determined by the best weighting function between fixed exponential and adaptive exponential. The optimum weighting function with a minimum cross-validation (CV) value of 1.7317×106 is adaptive exponential. Based on GWPR parameter testing, factors that affect DHF are local and diverse in each 10 regencies/municipalities in East Kalimantan Province. These factors are population density, number of health facilities, percentage of proper sanitation use in the household, percentage of household with qualified drinking water sources, and percentage of health services. The coefficient of determination of the GWPR model obtains a higher value than the FEM, which is 95.33%. Based on the measurement of goodness using the coefficient of determination value, it can be concluded that GWPR is the best method to model the DHF data rather than the FEM.