Model Geographically Weighted Poisson Regression (GWPR) dengan Fungsi Pembobot Adaptive Gaussian

Ridhawati, Ridhawati
Suyitno, Suyitno
Wasono, Wasono

Eksponensial
Vol 12 No 2 (2021)

The Geographically Weighted Poisson Regression (GWPR) Model is a regression model developed from Poisson regression or a local form of Poisson regression. The GWPR model generates a local model parameter estimator at each observation location where the data is collected and assumes the data is Poisson distributed. The estimation of GWPR model parameters uses the Adaptive Gaussian weighting function by determining the optimum bandwidth using GCV criteria. Based on the GWPR model, it is found that the factors that influence the maternal mortality rate (MMR) data in 24 districts (cities) of East Kalimantan and West Kalimantan are the percentage of pregnant women receiving Fe3 tablets, pregnant women with obstetric complications and the number of hospitals. These three variables produce four groups of GWPR model. Based on the GCV value, it is obtained that the best model is the GWPR model because it has the smallest GCV value.