Title: On linear generating processes for joint distributions of binary variables having an undirected Markov graph structure
Authors: Nanny Wermuth - Chalmers University of Technology (Sweden) [presenting]
Giovanni Maria Marchetti - University of Florence (Italy)
Abstract: In general, linear relations among binary variables are not suitable to capture conditional dependences or independences among categorical variables. But, we can give necessary and sufficient conditions on undirected Markov graphs so that the associated joint distributions of standardised binary variables have recursive, linear generating processes. Surprisingly, graphs of chordless cycles belong to the class. The attractive properties of these models and of their parameters are summarised.