Abstract

Cause-effect relationships are of primary interest in many fields of science. We consider the problem of estimating such relationships from observational data, that is, from data obtained by observing the system of interest without subjecting it to interventions or perturbations. We discuss that, under some assumptions, it is possible to consistently estimate bounds on causal effects from such data, even in high-dimensional settings where the number of variables is much larger than the sample size. We present an experimental validation of our method, using a challenging high-dimensional yeast gene expression data set, where we could indeed find the strongest causal effects between genes. An important application of this new type of statistical inference is that it offers useful strategies for the prioritization of experiments.

http://stat.ethz.ch/~maathuis/