A Bayesian nonparametric test for conditional independence
This provides a new tool for researchers in statistics and causal inference, though it appears incremental as it builds on existing Bayesian nonparametric approaches.
The paper tackles the problem of quantifying evidence for conditional dependence or independence between two variables given a third, introducing a Bayesian nonparametric method that uses Polya tree priors to account for uncertainty in underlying distributions, resulting in a symmetric probability measure advantageous for causal discovery.
This article introduces a Bayesian nonparametric method for quantifying the relative evidence in a dataset in favour of the dependence or independence of two variables conditional on a third. The approach uses Polya tree priors on spaces of conditional probability densities, accounting for uncertainty in the form of the underlying distributions in a nonparametric way. The Bayesian perspective provides an inherently symmetric probability measure of conditional dependence or independence, a feature particularly advantageous in causal discovery and not employed in existing procedures of this type.