Hyper-Poisson Model for Overdispersed and Underdispersed Count Data

Abstract

The Poisson model is commonly used for modelling count data. However, it has a limitation, namely the equality between the mean and variance (equidispersion) of the data to be modeled. Unfortunately, overdispersion (variance greater than the mean) and underdispersion (variance smaller than the mean) are more often to be found in real cases. Therefore, different models need to be used to handle data with these cases. The hyper-Poisson model is one model that can be used to handle overdispersion or underdispersion cases flexibly. This paper describes the hyper-Poisson model and its application on overdispersed and underdispersed count data.