MENGATASI OVERDISPERSI DENGAN REGRESI BINOMIAL NEGATIF PADA ANGKA KEMATIAN IBU DI KOTA BANDUNG

Abstract

The maternal mortality rate in the city of Bandung is still a concern for the government, even though various health programs have been held to handle it. The very slight reduction in maternal mortality is a reason for further research to look for factors that have a significant effect. The data on maternal mortality cases usually contain a lot of zeros and follow the Poisson distribution so that they are solved with a Poisson regression model, however the model formed cannot be used because the model shows overdispersion with a deviation value of more than one. Therefore, to overcome this problem, negative binomial regression is used as a solution. This negative binomial regression model produces three predictor variables out of seven variables that have a significant effect on maternal mortality in the city of Bandung including pregnant women receiving FE1 (30 tablets), deliveries assisted by health personnel and postpartum service coverage. Then tested the goodness of the model from the negative binomial regression model by looking at the AIC value. The true negative binomial regression model is better because the AIC value is 109.4 which is smaller than 121.65 which is the AIC value of the Poisson regression model.