DeepAI AI Chat
Log In Sign Up

Cause-Effect Inference in Location-Scale Noise Models: Maximum Likelihood vs. Independence Testing

by   Xiangyu Sun, et al.

Location-scale noise models (LSNMs) are a class of heteroscedastic structural causal models with wide applicability, closely related to affine flow models. Recent likelihood-based methods designed for LSNMs that infer cause-effect relationships achieve state-of-the-art accuracy, when their assumptions are satisfied concerning the noise distributions. However, under misspecification their accuracy deteriorates sharply, especially when the conditional variance in the anti-causal direction is smaller than that in the causal direction. In this paper, we demonstrate the misspecification problem and analyze why and when it occurs. We show that residual independence testing is much more robust to misspecification than likelihood-based cause-effect inference. Our empirical evaluation includes 580 synthetic and 99 real-world datasets.


page 22

page 27

page 28

page 29


Distinguishing Cause from Effect Based on Exogeneity

Recent developments in structural equation modeling have produced severa...

On the Identifiability and Estimation of Causal Location-Scale Noise Models

We study the class of location-scale or heteroscedastic noise models (LS...

Causal Identification with Additive Noise Models: Quantifying the Effect of Noise

In recent years, a lot of research has been conducted within the area of...

Causal Mosaic: Cause-Effect Inference via Nonlinear ICA and Ensemble Method

We address the problem of distinguishing cause from effect in bivariate ...

Telling Cause from Effect using MDL-based Local and Global Regression

We consider the fundamental problem of inferring the causal direction be...

Causal Discovery by Kernel Intrinsic Invariance Measure

Reasoning based on causality, instead of association has been considered...