[From Rick Marken (2010.08.31.0850)]
There was a table created by Richard Kennaway (or Bill, based on Richard’s equations) that (I think) the probability of correctly predicting an individual’s Y score give the correlation between X and Y scores. Could one of you (Richard or Bill) please post it when you get a chance.
Thanks
Best
Rick
–
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com
[From Richard Kennaway (2010.08.31.1709 BST)]
[From Rick Marken (2010.08.31.0850)]
There was a table created by Richard Kennaway (or Bill, based on Richard's equations) that (I think) the probability of correctly predicting an individual's Y score give the correlation between X and Y scores. Could one of you (Richard or Bill) please post it when you get a chance.
I have this on a web page at http://www2.cmp.uea.ac.uk/~jrk/distribution/correlations/corr.html
See the first two columns of the table in the section "Binary classification and screening", which give the probability of correctly guessing just the sign of Y given the value of X and the correlation. The table following gives the probability of correctly predicting the decile of Y.
There's also a full (but unpublished) paper at http://www2.cmp.uea.ac.uk/~jrk/distribution/correlationinfo.pdf
I'm currently trying to write anbother paper (I know, I should get the correlations one published) on the impossibility of predicting causation from correlation in dynamical systems. (Some of this is contained in the version of the correlations paper above, but I think it's sufficiently different to split off as a separate paper.) There's a large body of work on predicting causation from correlation: the major works are Judea Pearl's book "Causality" and Spirtes, Glymour, and Scheines' book "Causation, Prediction, and Search". However, all of that work makes certain assumptions about the systems being studied that either rule out all circular patterns of causation, or rule out dynamic relationships (i.e. where one variable influences not the value of another variable, but its rate of change). The paper is intended to demonstrate that real and very commonplace physical systems, especially control systems, can robustly violate these assumptions, and than when this is so, the patterns of correlations among the variables can be pretty much independent of the patterns of causal relationships.
--
Richard Kennaway, jrk@cmp.uea.ac.uk, Richard Kennaway
School of Computing Sciences,
University of East Anglia, Norwich NR4 7TJ, U.K.
[From Rick Marken (2010.08.31.1315)]
Thanks Richard. Bookmarked!
I’ll send you my paper on causation (the one for which I stole your title: When Causation Does Not Imply Correlation) if it ever gets published (I don’t want to send it now because it keeps morphing as I try to get it past different journal reviewers;-).
Best
Rick
On Tue, Aug 31, 2010 at 9:12 AM, Richard Kennaway jrk@cmp.uea.ac.uk wrote:
[From Richard Kennaway (2010.08.31.1709 BST)]
[From Rick Marken (2010.08.31.0850)]
There was a table created by Richard Kennaway (or Bill, based on Richard’s equations) that (I think) the probability of correctly predicting an individual’s Y score give the correlation between X and Y scores. Could one of you (Richard or Bill) please post it when you get a chance.
I have this on a web page at http://www2.cmp.uea.ac.uk/~jrk/distribution/correlations/corr.html
See the first two columns of the table in the section “Binary classification and screening”, which give the probability of correctly guessing just the sign of Y given the value of X and the correlation. The table following gives the probability of correctly predicting the decile of Y.
There’s also a full (but unpublished) paper at http://www2.cmp.uea.ac.uk/~jrk/distribution/correlationinfo.pdf
I’m currently trying to write anbother paper (I know, I should get the correlations one published) on the impossibility of predicting causation from correlation in dynamical systems. (Some of this is contained in the version of the correlations paper above, but I think it’s sufficiently different to split off as a separate paper.) There’s a large body of work on predicting causation from correlation: the major works are Judea Pearl’s book “Causality” and Spirtes, Glymour, and Scheines’ book “Causation, Prediction, and Search”. However, all of that work makes certain assumptions about the systems being studied that either rule out all circular patterns of causation, or rule out dynamic relationships (i.e. where one variable influences not the value of another variable, but its rate of change). The paper is intended to demonstrate that real and very commonplace physical systems, especially control systems, can robustly violate these assumptions, and than when this is so, the patterns of correlations among the variables can be pretty much independent of the patterns of causal relationships.
–
Richard Kennaway, jrk@cmp.uea.ac.uk, http://www.cmp.uea.ac.uk/~jrk/
School of Computing Sciences,
University of East Anglia, Norwich NR4 7TJ, U.K.
–
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com