This paper presents a review about the theory of regression analysis based on Jorgensen`s dispersion models, which extends the classical theory of generalized linear models (GLMs). Dispersion models provide a rich class of parametric models useful to conduct regression analysis of non-normal data, which contain all the error distributions considered in the classical GLMs. One example used to illustrate the extension is the simplex GLM for continuous proportional data. Details concerning the maximum likelihood estimation, are presented, accompanied with some numerical results.