网站大量收购独家精品文档,联系QQ:2885784924

A theoretical study of Y structures for causal discovery.pdf

A theoretical study of Y structures for causal discovery.pdf

  1. 1、本文档共10页,可阅读全部内容。
  2. 2、原创力文档(book118)网站文档一经付费(服务费),不意味着购买了该文档的版权,仅供个人/单位学习、研究之用,不得用于商业用途,未经授权,严禁复制、发行、汇编、翻译或者网络传播等,侵权必究。
  3. 3、本站所有内容均由合作方或网友上传,本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺!文档内容仅供研究参考,付费前请自行鉴别。如您付费,意味着您自己接受本站规则且自行承担风险,本站不退款、不进行额外附加服务;查看《如何避免下载的几个坑》。如果您已付费下载过本站文档,您可以点击 这里二次下载
  4. 4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等,请点击“版权申诉”(推荐),也可以打举报电话:400-050-0827(电话支持时间:9:00-18:30)。
查看更多
A theoretical study of Y structures for causal discovery

A theoretical study of Y structures for causal discovery Subramani Mani? mani@ Department of EECS University of Wisconsin-Milwaukee Milwaukee, WI 53211 Peter Spirtes ps7z@ Department of Philosophy Carnegie Mellon University Pittsburgh, PA 15213 Gregory F. Cooper gfc@ Center for Biomedical Informatics University of Pittsburgh Pittsburgh, PA 15213 Abstract Causal discovery from observational data in the presence of unobserved variables is chal- lenging. Identification of so-called Y sub- structures is a sufficient condition for ascer- taining some causal relations in the large sample limit, without the assumption of no hidden common causes. An example of a Y substructure is A → C, B → C, C → D. This paper describes the first asymptotically reliable and computationally feasible score- based search for discrete Y structures that does not assume that there are no unobserved common causes. For any parameterization of a directed acyclic graph (DAG) that has scores with the property that any DAG that can represent the distribution beats any DAG that can’t, and for two DAGs that represent the distribution, if one has fewer parame- ters than the other, the one with the fewest parameter wins. In this framework there is no need to assign scores to causal structures with unobserved common causes. The paper also describes how the existence of a Y struc- ture shows the presence of an unconfounded causal relation, without assuming that there are no hidden common causes. 1 Introduction Discovering causal relationships from observational data is challenging due to the presence of observed con- founders1, particularly, hidden (latent) confounders. ?Currently at the Department of Biomedical Informat- ics, Vanderbilt University, Nashville, TN 37232-8340. 1A node W is said to be a confounder of nodes A and B if there is a directed path from W to A and a directed path from W to B that does not traverse A. If W is observed, it is said to be a measured confounder, otherwise it is a hidden c

文档评论(0)

l215322 + 关注
实名认证
内容提供者

该用户很懒,什么也没介绍

1亿VIP精品文档

相关文档