Modeling bivariate Poisson regression for maternal and infant mortality in Central Java

Abstract

Regression analysis is an analysis of the relationship between dependent variables and independent variables. If the dependent variable is a count data and has Poisson distribution, a Poisson regression model is developed. At present the Poisson regression model is developed by the Bivariate Poisson Regression model. This model is estimated by using the Expectation Maximization (EM) algorithm. The Bivariate Poisson model produces 3 models, namely the variances in the form of equations, constants and zero. In this study, we would like to applied bivariate poisson regression to model maternal and infant mortality in Central Java. From this model, we know the factors that impact the increasing of maternal and infant mortality. The obtained of result study was that in the Bivariate Poisson, the best model is the second model which assumes that covariance is an equation. In this model, variables that significantly influence infant mortality are the percentage of pregnant women implementing K1 (X1), percentage of pregnant women implementing K4 (X2), percentage of pregnant women who received Fe3 tablets (X3), percentage of birth helped by health personnel (X4), percentage of obstretical complication handled (X5), percentage of childbirth women that have puerperal health service (X7) and percentage of household with clean and healthy behavior (X8) and there are no variables that influence maternal mortality. The best model has AIC value in the amount of 1114.5763.