Modeling Count Data Using Poisson and Negative Binomial Regression

Count data, such as the number of traffic accidents, hospitalizations, or criminal offenses, are common in applied statistics. Classical linear regression is not suitable for this type of data because it does not respect their distributional properties. Instead, Poisson and negative binomial regression models are commonly used. This thesis focuses on explaining and comparing these two models using real data. Special attention is paid to the problem of overdispersion, the use of offsets to account for different levels of exposure, and the interpretation of model coefficients. The results are discussed with an emphasis on practical interpretation and model suitability.