Nanny Wermuth
Mathematical Sciences
Göteborg University


Types of graphical Markov Models, illustrated
with symmetric binary variables.

Abstract

A wealth of different types of graphical Markov models
has been developed by now, with multivariate
regression chains and their subclass of triangular systems
being  most suitable for  modelling development in
joint and single responses over time.

Symmetric binary variables arise for instance when
continuous variables are categorized into two groups
with the cut-off value being the median. However, their
attractiveness for demonstrating similarities and differences
between simple models that are typical for different
model classes stems from closed form representations, also
after conditioning or marginalising.