Examining the feasibility of the negative binomial model in characterizing microbial abundance variation

Abstract

The simplicity of a linear model makes it a powerful tool for studying natural phenomena. However, often the assumptions are too limited. In statistical ecology, for instance, we frequently encounter situations where variation in species abundances fluctuate in a non-linear manner and exhibit properties such as heteroscedasticity that are not captured by standard linear models. In this thesis, I investigate the feasibility of the negative binomial model, as a tool in statistical ecology. The negative binomial has many desirable properties in terms of modelling variation in species abundances. I discuss these properties and assess the performance of a specific type of the negative binomial model which is called the traditional negative binomial model in the context of microbial ecology, which is a rapidly emerging application area for statistical models. The analysis is based on openly available data sets from published literature. The thesis concludes by discussing the potentials of the negative binomial models and their challenges.