Geographically weighted bivariate Poisson inverse Gaussian regression modeling with the Berndt-Hall-Hall-Hausman algorithm on maternal and neonatal mortality data

Abstract

Maternal and neonatal mortality rates in South Sulawesi remain higher than the national average across all provinces in Indonesia. This study aims to identify significant variables for each district/city in South Sulawesi, Indonesia. The data used was overdispersed in the two cases which correlated and distributed Poisson. The Gaussian Poisson Inverse Bivariate Regression can be used to solve the problem but cannot solve the problem of spatial heterogeneity. Spatial heterogeneity causes bias in the interpretation of results. The method to overcome this problem is the Geographycally Weighted Bivariate Poisson Inverse Gaussian method. The Berndt-Hall-Hall-Hausman algorithm is used in the parameter estimation of the GWBPIGR model. The Kernel functions used are Adaptive Bisquare, Adaptive Tricube, and Fixed Gaussian. Generalized Cross Validation (GCV) is used to select the optimal bandwidth. The results of this study show that  the Akaike Information Criterion (AIC) value in the GWBPIGR model with the Berndt-Hall-Hall-Hausman algorithm is better than that of BPIGR with the same algorithm.