Generalized estimating equations when the response variable has a Tweedie distribution: an application for multi-site rainfall modelling

Abstract

This dissertation will focus on modelling rainfall processes by using generalized estimating equations, when the assumed distribution of the response variable is from the Tweedie family of distributions. The Tweedie distribution allows the different components of rainfall to be modelled simultaneously. When the Tweedie distributions are incorporated into generalized estimating equation estimation techniques, not only can the non-independent structure of data be incorporate but also multiple rainfall sites can be modelled concurrently. This dissertation will attempt to demonstrate the potential benefits of using generalized estimating equations for modelling and interpreting historical rainfall records by the examination of of monthly rainfall models at Emerald, Toowoomba and Gatton. The suitability and predictability of the models presented in this dissertation will then be discussed to determine if generalized estimating equations provide a unique method of modelling rainfall.