Geographic Modeling on The Infant Mortality Rate In West Java

Abstract

In geographic modeling, global models such as ordinary linear regression (OLR) model theoretically it provides quite reliable local information if there is not any spatial diversity by region. In other words, OLR model cannot describe the relations between variables in heterogeneous difference of each region. This study will consider a model that will be used to estimate or predict the infant mortality rate in the several regencies / cities in West Java Province. Because the response variable observed in this study is count data which is assumed Poisson distributed, geographically weighted Poisson regression model (GWPR) is used. A better model is used to analyze the data of infant deaths in each regency / city in West Java based on the AIC value, GWPR model has the smallest value (compared to Poisson regression model), in which there is an interesting and important difference from each regency/city about the factors that significantly influence the Infant Mortality rate in each region.