Conditional distribution variability measures for causality detection

TL;DR

This paper introduces variability measures for conditional distributions to infer causal relationships, demonstrating their effectiveness in a challenge with a high AUC score and second place ranking.

Contribution

It proposes new variability measures for conditional distributions and combines them with statistical measures for improved causality detection.

Findings

Achieved an AUC score of 0.82 on the test database.

Ranked second in the ChaLearn cause-effect pair challenge.

Demonstrated the usefulness of variability measures in causal inference.

Abstract

In this paper we derive variability measures for the conditional probability distributions of a pair of random variables, and we study its application in the inference of causal-effect relationships. We also study the combination of the proposed measures with standard statistical measures in the the framework of the ChaLearn cause-effect pair challenge. The developed model obtains an AUC score of 0.82 on the final test database and ranked second in the challenge.