Local independence graphs for composable Markov processes.
Sonderforschungsbereich 386, Discussion Paper 158
The concept of local independence is used to define local independence graphs representing the dynamic dependence structure of several continuous time processes which jointly form a so-called composable Markov process. Specific properties of this new class of graphs are discussed such as the role of separating sets. Further insight is gained by considering possible extensions to the discrete time situation. It is shown that the latter case can be reduced to classical graphical interaction models.