Graphical Markov models: a tool for interpretation

David R. Cox
Nuffield College and Department of Statistics
University of Oxford, UK

Abstract

Graph-theoretical representations of multivariate systems were
introduced 80 years by Sewall Wright. In recent years they have
been developed in various contexts, sometimes largely as
computational devices. In this paper an approach aimed at the
interpretation of empirical data is developed from first
principles and some consequences set out. Matrix methods are
emphasized and the connection with discussions of causality
outlined.

The talk is based on joint work with Nanny Wermuth, Chalmers /
Gothenburg University.