Journal of Causal Inference (May 2023)

On the pitfalls of Gaussian likelihood scoring for causal discovery

  • Schultheiss Christoph,
  • Bühlmann Peter

DOI
https://doi.org/10.1515/jci-2022-0068
Journal volume & issue
Vol. 11, no. 1
pp. 689 – 96

Abstract

Read online

We consider likelihood score-based methods for causal discovery in structural causal models. In particular, we focus on Gaussian scoring and analyze the effect of model misspecification in terms of non-Gaussian error distribution. We present a surprising negative result for Gaussian likelihood scoring in combination with nonparametric regression methods.

Keywords