Introduction to Graphical Markov Models

Contact: wermuth@chalmers.se Course number: None, but may be taken instead of "Multivariate Analysis" [CTH-TMS041 or GU-MSA200] Class times and room: Tuesday, 27. Oct 2009 until Wednesday, 16. Dec 2009 Tuesdays 16:00-17:45 in MVH11 Wednesdays 9:15-11:00 in MVH11 On Tuesday, 3. Nov, there will be a guest lecture by D.R. Cox, Nuffield College. Background: Graphical Markov models have several properties that distinguish them from traditional multivariate statistical model classes and that needed explication over the years. The models permit graphical representations, where the graphs capture an independence structure of interest. Available substantive prior knowledge can be integrated via local specifications. Many model implications can be derived for instance after intervention or after marginalizing over some of the variables and conditioning on others. The models can be implemented under key types of different distributional assumptions, for instance for multivariate exponential family distributions. Password protected section (if it does not work please contact webmaster) Includes: reading, data and exercises.