{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/conditional-independence-testing-based-on-a","title":"Conditional independence testing based on a nearest-neighbor estimator of conditional mutual information","arxiv_id":"1709.01447","date":"2017-09-05","proceeding":null,"authors":["Jakob Runge"],"abstract":"Conditional independence testing is a fundamental problem underlying causal\ndiscovery and a particularly challenging task in the presence of nonlinear and\nhigh-dimensional dependencies. Here a fully non-parametric test for continuous\ndata based on conditional mutual information combined with a local permutation\nscheme is presented. Through a nearest neighbor approach, the test efficiently\nadapts also to non-smooth distributions due to strongly nonlinear dependencies.\nNumerical experiments demonstrate that the test reliably simulates the null\ndistribution even for small sample sizes and with high-dimensional conditioning\nsets. The test is better calibrated than kernel-based tests utilizing an\nanalytical approximation of the null distribution, especially for non-smooth\ndensities, and reaches the same or higher power levels. Combining the local\npermutation scheme with the kernel tests leads to better calibration, but\nsuffers in power. For smaller sample sizes and lower dimensions, the test is\nfaster than random fourier feature-based kernel tests if the permutation scheme\nis (embarrassingly) parallelized, but the runtime increases more sharply with\nsample size and dimensionality. Thus, more theoretical research to analytically\napproximate the null distribution and speed up the estimation for larger sample\nsizes is desirable.","url_abs":"http://arxiv.org/abs/1709.01447v1","url_pdf":"http://arxiv.org/pdf/1709.01447v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"conditional-independence-testing-based-on-a","repo_url":"https://github.com/jakobrunge/tigramite","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok","spdx":"GPL-3.0"}}],"tasks":[{"task_slug":"causal-discovery","task_name":"Causal Discovery"}],"methods":[{"method_slug":"speed","method_name":"SPEED"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1709.01447","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}