Robust regression with MM-estimator for modelling the number maternal mortality of pregnancy in Central Java, Indonesia

Abstract

Regression is a method of statistics analysis for modeling between the dependent variable to independent variables. Estimation in the regression model uses the Ordinary Least Square (OLS) method. In the regression model, several assumptions are required. One of the reasons for the violation of the assumption is that there are outliers in the data. There are several methods that can be used to overcome outliers, one of which is using the Robust MM-estimator regression. The MM-estimator has high efficiency when the error is normally distributed and also has a high breakdown value. To implement the SDG’s program, one program is health. It has objectiveness to reduce maternal mortality rate. Maternal mortality consist of pregnancy, give of birth and postpartum mother. In this study the data used was the number of maternal mortality of pregnancy in Central Java in 2019 as the dependent variable, while the independent variables used included the percentage of pregnant women consuming Fe3 tablet (X1), percentage of households that have a clean and healthy lifestyle (X2); percentage of pregnant women who made the first visit (X3); The results obtained in multiple regression modeling produce R square value of 0.222679 and Mean Square Error (MSE) of 6.894871. The assumptions in the regression model unfulfilled, may be caused by outlier in data. In MM robust regression modeling, the R square value is 0.2695, and the MSE value is 4.989. It shows that robust regression modeling with MM estimator is better than multiple regression.