Parameter Estimation and Statistical Test of Mixed Geographically Weighted Bivariate Weibull Regression (MGWBWR) : On the Cases of Infant Mortality and Maternal Mortality Rate in East Java 2016

In this research, an estimation and statistic test of a mixed geographically weighted bivariate model of Weibull regression (MGWBWR) to determine the factors that influence dependent variables will be discussed. This study has objectives to review the parameter estimation and test the significance of the MGWBWR model also will be applied on data problems of infant mortality and maternal mortality in East Java. The results showed that parameter estimation MGWBWR model with consistent and efficient Maximum Likelihood Estimation (MLE) method resulted a non-close form equation because the equations generated from the first derivative still contained parameters so it requires the iteration of Berndt, Hall, Hall, Hausman (BHHH) to determine parameter estimation. Based on data analysis of infant mortality rate and maternal mortality in East Java in 2016 using MGWBWR methods, variable influencing to IMR in Blitar Regency are Percentage of live-behavioral household clean and healthy, Percentage Antenatal Care (ANC) of Pregnant Women and variables affecting MMR in Blitar Regency are Ratio of number of health facilities on 100,000 population, Percentage of women in East Java who married under 17 years, Percentage of poor, Percentage Antenatal Care (ANC) of Pregnant Women.