Zero Inflation Model for Food Poisoning index

Abstract

In this thesis, we review the event count data analysis methods. In particular, we focus on the zero-inflated regression and time series models in contrast to the well-known GLM regression model. The event count regression models including Poisson regression and negative binomial regression are introduced. Then, we analyze the food poisoning data using the zero-inflated models which consider mixed probability distributions. It is well known that the zero-inflated models can catch the special characteristics such as the over-dispersion and zero-inflation of the real life data. We compare the performances of the zero-inflated models with the other models using Vuong statistic and propose new food-poisoning index based on the zero-inflated model.