A final project carried out at the African Institute for Mathematical Sciences, AIMS.
ABSTRACT
In statistics, there are two approaches to model fitting. This study applied both approaches, the Classical and Bayesian (using Markov chain Monte Carlo), in fitting Generalized Linear Models, GLM. We considered fitting the GLM using three distributions from the exponential family of distribution; the Normal, Binomial and Poisson. Results showed that both approaches used gave quite similar model parameters’ values. The Mean squared error, Akaike Information Criteria and Bayesian Information Criteria were used as tools to select the best model distribution. Based on the mortality data used, the Binomial and Poisson distributions are most appropriate as the model distributions as they both had the least MSE.