Exponential Dispersion Models are a two-parameter distribution probability models utilized in Generalized Linear Models. An a priori distribution of this family is postulated for the latent model and for the observation model in a Kalman Filter framework. A Bayesian estimation is performed to obtain an a posteriori distribution for the latent and observation states. In some cases, known distributions are found and we present some of those cases and an algorithm for performing
extending Kalman Filtering technique to non-Gaussian distributions.