English

An inferential measure of dependence between two systems using Bayesian model comparison

Machine Learning 2024-12-11 v2 Machine Learning Quantitative Methods

Abstract

We propose to quantify dependence between two systems XX and YY in a dataset DD based on the Bayesian comparison of two models: one, H0H_0, of statistical independence and another one, H1H_1, of dependence. In this framework, dependence between XX and YY in DD, denoted B(X,YD)B(X,Y|D), is quantified as P(H1D)P(H_1|D), the posterior probability for the model of dependence given DD, or any strictly increasing function thereof. It is therefore a measure of the evidence for dependence between XX and YY as modeled by H1H_1 and observed in DD. We review several statistical models and reconsider standard results in the light of B(X,YD)B(X,Y|D) as a measure of dependence. Using simulations, we focus on two specific issues: the effect of noise and the behavior of B(X,YD)B(X,Y|D) when H1H_1 has a parameter coding for the intensity of dependence. We then derive some general properties of B(X,YD)B(X,Y|D), showing that it quantifies the information contained in DD in favor of H1H_1 versus H0H_0. While some of these properties are typical of what is expected from a valid measure of dependence, others are novel and naturally appear as desired features for specific measures of dependence, which we call inferential. We finally put these results in perspective; in particular, we discuss the consequences of using the Bayesian framework as well as the similarities and differences between B(X,YD)B(X,Y|D) and mutual information.

Keywords

Cite

@article{arxiv.2412.06478,
  title  = {An inferential measure of dependence between two systems using Bayesian model comparison},
  author = {Guillaume Marrelec and Alain Giron},
  journal= {arXiv preprint arXiv:2412.06478},
  year   = {2024}
}

Comments

To be published in IEEE Transaction on Systems, Man, and Cybernetics: Systems

R2 v1 2026-06-28T20:27:52.248Z