Predicting Causal Direction from Dependency Features
Description
The relationship between statistical dependency and causality lies
at the heart of all statistical approaches to causal inference. The D2C
package implements a supervised machine learning approach to infer the
existence of a directed causal link between two variables in multivariate
settings with n>2 variables. The approach relies on the asymmetry of some
conditional (in)dependence relations between the members of the Markov
blankets of two variables causally connected. The D2C algorithm predicts
the existence of a direct causal link between two variables in a
multivariate setting by (i) creating a set of of features of the
relationship based on asymmetric descriptors of the multivariate dependency
and (ii) using a classifier to learn a mapping between the features and the
presence of a causal link