{ "id": 4138, "title": "Formalism", "url": "/competitions/cause-effect-pairs/discussion/4138", "author_name": "", "post_date": "2013-03-29T22:30:17.670Z", "votes": null, "comment_count": 1, "views": 1365, "content": "

I wonder about the definition of "A-B" and "A|B" (e.g. in the details column of CEdata_train_target.csv). 

\r\n

Can somebody help me out?

", "messages": [ { "id": "21836", "postDate": "03/29/2013 22:30:17", "content": "

I wonder about the definition of "A-B" and "A|B" (e.g. in the details column of CEdata_train_target.csv). 

\r\n

Can somebody help me out?

", "rawMarkdown": "", "votes": null }, { "id": "21840", "postDate": "03/29/2013 23:01:30", "content": "

A-B means that the two variables are dependent, but NOT causally related, i.e., the dependency is explained by the presence of a common cause "Z". This is sometimes represented as A<-Z->B or A<->B. 

\r\n

A-B does NOT mean that we have both A<-B and A->B (cycles). To our knowledge, there is no instance of cycles in our data.

\r\n

A|B means that A and B are independent (not related at all whatsoever). This should be "easy" to detect with a stastistical independence test (with some uncertainty when there are few data points).

", "rawMarkdown": "", "votes": null } ], "comments": [ { "id": 21840, "author_name": "iguyon", "author_url": "", "post_date": "03/29/2013 23:01:30", "content": "

A-B means that the two variables are dependent, but NOT causally related, i.e., the dependency is explained by the presence of a common cause "Z". This is sometimes represented as A<-Z->B or A<->B. 

\r\n

A-B does NOT mean that we have both A<-B and A->B (cycles). To our knowledge, there is no instance of cycles in our data.

\r\n

A|B means that A and B are independent (not related at all whatsoever). This should be "easy" to detect with a stastistical independence test (with some uncertainty when there are few data points).

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "21836": "", "21840": "" }, "source": "meta" }