Graphical dynamic linear models: specification, use and graphical transformations Original Research Article

In this work, we propose a dynamic graphical model as a tool for Bayesian inference and forecasting in dynamic systems described by a series which is dependent on a state vector evolving according to a Markovian law. We build sequential algorithms for the probabilities propagation. This sequentiality turns out to be represented by the dynamic graphical structure after carrying out several goal-oriented sequential graphical transformations.