Comparison of Two Weighting Functions in Geographically Weighted Zero-Inflated Poisson Regression on Filariasis Data

Abstract

Spatial effects are factors to consider in modeling spatial data. These spatial effects can be spatial dependencies and spatial heterogeneity. The purposes of this study are: (1) to form Geographically Weighted Zero-Inflated Poisson (GWZIP) regression model to overcome the problem of spatial heterogeneity and the big enough proportion of zero-inflation in Filariasis case; and (2) to find the best weighting function between the fixed Gaussian kernel and the fixed Bi-square kernel based on the deviance of model. This study uses secondary data covering 35 districts in Central Java Province. The results of this study indicate that there are spatial heterogeneity and the 60% proportion of zero-inflation in Filariasis data. Based on the value of deviance model, it is known that the GWZIP model using fixed Gaussian kernel is better than the GWZIP model using fixed Bi-square kernel.