Infant Mortality Case: An Application of Negative Binomial Regression in order to Overcome Overdispersion in Poisson Regression

Infant mortality is an indicator to determine the degree of public health. Infant mortality is death that occurs in the period from birth to before the age of one. The high rate of infant mortality indicates that the quality of public health services is not optimal. The number of infant deaths is an example of count data that follows a Poisson distribution, so it can be analyzed using Poisson Regression. The assumption that must be met when using this method is the equidispersion or variance of the response variable is equal to mean. However, this condition rarely occurs because usually the counted data has a greater variance than the mean or it is called overdispersion. One way to solve this problem is to use the Negative Binomial Regression method. The data used in this study is the case of infant mortality in the city of Padang. First, we model the data using Poisson Regression, then we check the assumption, if there is overdispersion, we handle it by modeling the data with Negative Binomial Regression. The results showed that the equidispersion assumption could not be met so that the data was modeled with Negative Binomial Regression