An Asymptotic Test for Conditional Independence Using Analytic Kernel Embeddings

Abstract

We propose a new conditional dependence measure and a statistical test for conditional independence. The measure is based on the difference between analytic kernel embeddings of two well-suited distributions evaluated at a finite set of locations. We obtain its asymptotic distribution under the null hypothesis of conditional independence and design a consistent statistical test from it. We conduct a series of experiments showing that our new test outperforms state-of-the-art methods both in terms of type-I and type-II errors even in the high dimensional setting.

Cite

Text

Scetbon et al. "An Asymptotic Test for Conditional Independence Using Analytic Kernel Embeddings." International Conference on Machine Learning, 2022.

Markdown

[Scetbon et al. "An Asymptotic Test for Conditional Independence Using Analytic Kernel Embeddings." International Conference on Machine Learning, 2022.](https://mlanthology.org/icml/2022/scetbon2022icml-asymptotic/)

BibTeX

@inproceedings{scetbon2022icml-asymptotic,
  title     = {{An Asymptotic Test for Conditional Independence Using Analytic Kernel Embeddings}},
  author    = {Scetbon, Meyer and Meunier, Laurent and Romano, Yaniv},
  booktitle = {International Conference on Machine Learning},
  year      = {2022},
  pages     = {19328-19346},
  volume    = {162},
  url       = {https://mlanthology.org/icml/2022/scetbon2022icml-asymptotic/}
}