ABSTRACT
This chapter starts from Bayesian inference in any discrete model and discusses conditional independence models represented by directed acyclic graphs (DAGs). It gives a very abbreviated overview of the vast field of statistical inference. The chapter focuses in particular on modelling from a Bayesian perspective and the use of graphical models. It demonstrates how various implicit conditional independences can be read from a Chain Event Graph (CEG) that gives us insight over some of the global properties of its associated statistical model. The chapter suggests that the larger the sample size of the data the closer is the posterior estimate to the sample proportion, so in this case maximum likelihood estimation is asymptotically the same as the Bayesian approach. It considers making inference about a probability mass function in the situation where any such mass function is a plausible one over the space of all such functions on a given set of atoms.
