Geographically Weighted Ordinal Logistic Regression Model

Abstract

Regression analysis is a statistical analysis that is used to model the relationship between the response variable with one or more predictor variables. Ordinal logistic regression is used to model the relationship between the response variable which is categorical and ordinal scale with one or more predictor variables. If each of the regression coefficients depends on the geographic location where the data is observed then it is used the Geographically Weighted Ordinal Logistic Regression (GWOLR) model. Estimation of parameters GWOLR model can use weighted maximum likelihood method, which is by giving a different geographic weighting for each location on the ln-likelihood function. In this study, the type of weighting function used is Gaussian kernel function. GWOLR model applications to model the village vulnerability of dengue hemorrhagic disease in Lamongan 2009 showed that the GWOLR model is not significantly different with ordinal logistic regression model. But in this study, GWOLR model uses a Gaussian kernel weighting function has the accuracy of classification which is greater than an ordinal logistic regression model. Factors that significantly affect village vulnerability of dengue hemorrhagic disease in Lamongan 2009 based on GWOLR model are population density (X1), the distance from the village to the nearest health center (X3) and the distance from the village to the district capital (X6).