Kajian Overdispersi pada Regresi Poisson Menggunakan Semiparametrik Zero-Inflated Poisson

Abstract

One of the causes of overdispersion is that there are many zero observations on the response variable which can be detected by the dispersion value which is the ratio of deviance and degrees of freedom. The method to overcome this problem is the Zero-Inflated Poisson (ZIP). Semiparametric approach be used as an alternative model because it has parametric and nonparametric component so it has a high degree of flexibility. Absolute of Relative Bias (ARB) is used to determine the level of accuracy of parameter estimators and Root of Mean Square of Error (RMSE) is used to determine the goodness of fit of the model. Results show that the ARB smallest value from simulation data was found in ZIP for overdispersion while Poisson regression for nonoverdispersion. The smallest RMSE was found on semiparametric model that show that semiparametric model was more suitable for mixed data. On the semiparametric ZIP model for the maternal mortality rate in East Java application data, which had overdispersion indicated by the dispersion values is greater than one. This semiparametric ZIP model also has the smallest RMSE values so that it can be said to be the best model. Variable which affect the number of maternal mortality in East Java based on the semiparametric ZIP model is X1 (visits of pregnant women K1).