Thanks to the work of Pearl (2000) and Spirtes, Glymour, and Scheines (1993), there has been an explosion of research in causal inference over the last two decades. Current research is both theoretical and practical: new algorithms are being developed under increasing weak assumptions about causation, and automated causal discovery procedures, which infer causal structure (as represented by graphical structure) from statistical data, have been successfully applied in medical research and the social sciences. This course is an introduction to traditional "constraint-based" algorithms, which infer causal structure from patterns of conditional independence among the variables. We will then discuss how these algo-rithms are being extended to data sets with large numbers of variables.
Topics will include: