Hurdle Regression Modelling on the Number of Deaths from Chronic Filariasis Cases in Indonesia

Abstract

Poisson regression is one of the model to explain the functional relationship between response variable in the form of count and predictor variable. An important assumption in Poisson Regression analysis is equidispersion. In certain cases, where response variable consists of too many zeros, causing the variance to be greater than the mean or called overdispersion that can be overcomed by the Hurdle model. Filariasis disease is caused by filaria worm that led to swelling of the limbs in humans. One province in Indonesia, Papua Barat, reported a quite high death of chronic filariasis cases with a death rate of 459 people. The Hurdle regression model is appropriately to model the number of cases of chronic filariasis death in Indonesia since the data contains overdispersion. This study will compared two regression Hurdle models, namely the Hurdle Poisson regression and Hurdle Negative Binomial Regression. The results showed that the Negative Binomial Hurdle regression model was better than that of the Hurdle Poisson regression model in modeling cases of filariasis in Indonesia with AIC value of 213.263. Based on logit model of Negative Binomial Regression, the percentage of households that have access to proper sanitation (𝑋 5 ) has a significant effect on the number of cases of death from chronic filariasis in Indonesia.