ParceLiNGAM: A causal ordering method robust against latent confounders

TL;DR

ParceLiNGAM is a new causal ordering algorithm for linear non-Gaussian acyclic models that remains effective even when latent confounders violate some assumptions, demonstrated through artificial and brain imaging data.

Contribution

It introduces a robust method for causal ordering that detects latent confounders by testing independence and identifying unaffected variable subsets.

Findings

Effective in detecting causal order despite latent confounders

Performs well on artificial data

Shows promising results on simulated brain imaging data

Abstract

We consider learning a causal ordering of variables in a linear non-Gaussian acyclic model called LiNGAM. Several existing methods have been shown to consistently estimate a causal ordering assuming that all the model assumptions are correct. But, the estimation results could be distorted if some assumptions actually are violated. In this paper, we propose a new algorithm for learning causal orders that is robust against one typical violation of the model assumptions: latent confounders. The key idea is to detect latent confounders by testing independence between estimated external influences and find subsets (parcels) that include variables that are not affected by latent confounders. We demonstrate the effectiveness of our method using artificial data and simulated brain imaging data.