Study of Small Area Estimation on Overdispersion Data with the Zero-Inflated Poisson Regression

Small area estimation method is a method that is widely used to estimate the parameters of sub populations based on available survey data with very small sample sizes. Infant mortality data is a rare occurrence in a certain period of time so that small area estimation model is used with Poisson distribution. However, infant mortality data has a large number of zero value in each survey unit, so that the data indicating the existence of overdispersion that will certainly have an effect on the estimation process, then the proposed model is small area estimation with zero-inflated Poisson. This study compares the small area estimation with Poisson regression and small area estimation with zero-inflated Poisson regression. The goodness of the model is tested through simulation and revealed that small area estimation with zero-inflated Poisson regression is better than small area estimation with Poisson regression. Implementation on infant mortality data in West Java also shows that small area estimation with zero-inflated Poisson regression is better than small area estimate with Poisson regression. It is indicated by smaller standard error of small area estimation with zero-inflated Poisson regression than small area estimation with Poisson regression model.