Geographically weighted generalized poisson regression model with the best kernel function in the case of the number of postpartum maternal mortality in east java

Abstract

This study proposes a Geographically Weighted Generalized Poisson Regression (GWGPR) model with the best kernel function to obtain a model of the number of postpartum maternal mortality in East Java Province in 2020 and determine the factors that affect the number of maternal postpartum mortality in East Java in 2020. The kernel functions used in this study are fixed bisquare kernel, fixed tricube kernel, and adaptive bisquare kernel. Optimum bandwidth selection using the Cross-Validation (CV) method. The results obtained the best model is the GWGPR model with a fixed bisquare kernel because it produces the smallest AIC value of 194.92. Variables significantly affecting the number of maternal postpartum mortality in East Java in 2020 vary in each district/city where there are three regional groups. The percentage of pregnant women who had a pregnancy visit K1, the percentage of pregnant women who had a pregnancy visit K4, the percentage of households receiving cash assistance, and the ratio of hospitals and health centres have a significant effect on Kabupaten Blitar, Mojokerto, Gresik, Bangkalan, Blitar City, Mojokerto City, Surabaya City. While the five predictor variables together significantly affect districts/cities included in group 3, such as Ponorogo, Trenggalek, Tulungagung, Kediri, Malang, Lumajang, Jember, Banyuwangi, Bondowoso and so on. Some of the highlights of the proposed approach are:•Generalized Poisson regression model using the maximum likelihood estimation (MLE) method.•The kernel functions used in the Geographically Weighted Generalized Poisson Regression (GWGPR) model to determine bandwidth are fixed bisquare, fixed tricube, and adaptive bisquare kernel functions selected using the Cross Validation (CV) method.•The computation procedure is easy to implement.