Prediksi Jumlah Kasus Baru Kusta dengan Metode Geographically Weighted Poisson Regression (GWPR)

ABSTRACT Poisson regression was obtained from the Poisson distribution, which is a theoretical distribution associated with a discrete random variable count, where each event follows the Poisson distribution. Leprosy data in Buton is one example of the data count. The main problem of the Poisson regression is when applied to the spatial data, the heterogeneity will occur. Spatial heterogeneity arises because of the condition of data in each location is not the same, both in terms of geographical, socio-cultural and other things that lie behind them. One impact of the emergence of spatial heterogeneity is regression parameters are varying spatially, so as to solve the problems on data spatial, the spatial modellingis done. Spatial modeling is appropriate for use Geographically Weighted Poisson Regression (GPWR). This study aims to determine the best models on the number of new cases of leprosy in Buton District in 2013. Studies conducted a study of non-reactive or unobtrusive method. The experiment was conducted in Buton in Southeast Sulawesi province May-June 2014. Units of analysis in this study is the data new cases of leprosy in every district in Buton. The results showed Geographicaly Weighted Poisson Regression Model (GWPR) yields the smallest AIC value, so the best modeling for the number of new cases of leprosy in Buton is Geographicaly Weighted Poisson Regression Model (GWPR) than the model Poisson regression model.