Decision Making Based on Causal Graphs
Dietrich Eherler, Peter Kischka · 2001
Graphs have been used since quite long in statistics as tools to represent probabilistic dependencies. The origins can be traced back to first applications in path analysis (Wright 1921). During the last decade developments took place that involved an improved understanding of the relationships between graphs and probabilities, on the one hand, and graphs and causal inference on the other (e. g. Pearl 1988, Spirtes et al. 1993). Graphical modeling has emerged as a key component of the KDD (knowledge discovery in databases) process in the way of extraction and depiction of dependencies from data. Further, graphical modelling provides a powerful symbolic machinery for causal analysis e. g. the problem of covariate selection and the assessment of causal effects in the sense of the potential response framework. If a decision maker fixes the value of a treatment variable the effects of this decision can easily be computed using the causal graph associated with this model. These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.