Talk:Coefficient of multiple correlation

Copyright violation
Parts of this article (see http://www.visualstatistics.net) violate copyright of Cruise Scientific and were removed. For further information contact info@visualstatistics.net.
 * none of the article now violates any copyright. Rjensen 05:27, 29 March 2007 (UTC)

X is not a random variable
X is NOT viewed as a random variable in the linear regression, though correlation between X and Y exists. The latter is not the exactly defined correlation from Probability. —Preceding unsigned comment added by 89.96.249.182 (talk) 15:01, 1 May 2009 (UTC)

Wholesale copyright violation
Almost the entire article was word-for-word out of one of the external references,. On 11 November 2010, not realizing that it was plagiarized, I rewrote some poorly worded parts of it. Today I realized the plagiarism, and I have deleted both the intro and the first titled section, and have written a new intro.

The copying from the source occurred on 9 July 2005 by an anon who hasn't contributed since 2005. On 12 November 2006 the copyright violation was purged, and on 28 January 2007 the entire page was redirected to linear regression. But on 29 March 2007 everything was restored by someone who presumably didn't realize that it was verbatim.

It seems redundant to have an article "multiple correlation" when there are more complete articles coefficient of determination and linear regression, so maybe this page should be redirected to one of them. However, the key equation in this article is quite insightful and does not appear in those articles, so if this page is redirected then this equation and its discussion should be inserted there.

Thoughts, anyone? Duoduoduo (talk) 17:48, 21 November 2010 (UTC)

I delved into this some more. The copying was done from an online book by David J. Krus. The anon who copied it apparently now signs in as User:David_Cruise, and has received a lot of warnings on his talk page (up until October 2008) and has been rebuked for blanking parts of discussion pages that criticize his contributions -- see e.g. this edit. Apparently he hasn't edited since 13 August 2006. Duoduoduo (talk) 20:59, 21 November 2010 (UTC)

Formula is wrong?
There are some inconsistencies in the definition here, and I am guessing that the true value is supposed to be a scalar, not a matrix. Are we missing another 'c' on the right of the right-hand term? Shabbychef (talk) 18:55, 30 May 2012 (UTC)
 * Vandalism now fixed. Melcombe (talk) 07:53, 31 May 2012 (UTC)

Problem in the "Definition section"
This section does not provide a definition. It says the coefficient can be calculated by a given formula. The discipline of statistics is confounded by statements that confuse definitions from computational formulas, approximations, and interpretations. If there *is* a generally agreed up definition for the coefficient, please list it. If there is disagreement or confusion or multiple definitions which depend on a given context, please elucidate this. That would be very helpful. — Preceding unsigned comment added by Mrtweedles (talk • contribs) 14:11, 2 November 2012 (UTC)


 * Done. Thanks for the suggestion! Duoduoduo (talk) 16:39, 18 December 2012 (UTC)

Unclear meaning?
The first half of


 * Unlike the coefficient of determination in a regression involving just two variables, the coefficient of multiple correlation is not computationally commutative

was recently deleted on the grounds of its being unclear. Maybe I'm missing something, but it seems to me that this is both clear and useful in appreciating the contrasting information in the last half of the sentence: It says that the coefficient of determination in a regression involving just two variables (a special case of the coefficient of multiple correlation) is commutative -- that is, the R2 of regressing y on x is the same as that from regressing x on y (in each case it's the square of the correlation coefficient). Can we restore this (possibly in clarified form)? Duoduoduo (talk) 12:58, 5 April 2013 (UTC)


 * (i) "a regression involving just two variables" is easily interpreted as meaning two explanatory variables
 * (ii) "commutative" seems wrong as a possibility for multiple correlation as any reasonable notation must make a distinction between the (single) variable being predicted and the (one or more) explanatory variables.... so there is a non-symmetric stating point for any notional function representing the cofficient of multiple correlation.
 * (iii) "computationally" here seems misused, as it is the mathematics that count.
 * It may be best to wait until some decent notation is introduced, since this will allow the inclusion of many more important facts about multiple correlations. Melcombe (talk) 18:54, 5 April 2013 (UTC)

Multiple Correlation Coefficient
Can it be checked if the multiple correlation coefficient is represented as R2? My small research into such matters state that the multiple correlation coefficient is represented as just R, while the coefficient of determination is R2 — Preceding unsigned comment added by 220.245.17.242 (talk) 13:32, 20 May 2013 (UTC)


 * You're right -- thanks for catching that! I've corrected it in the article now. Duoduoduo (talk) 15:06, 20 May 2013 (UTC)

Regression?
Aren't "multiple correlation" and "regression" the same thing? There is a wiki page for each. Jfgrcar (talk) 19:11, 14 April 2014 (UTC)


 * Regression is a process for analyzing data. Multiple correlation is a quantitative feature of the data that can be found via regression. Loraof (talk) 20:00, 7 February 2015 (UTC)

Correct but unhelpful?
In order to be able to understand the introduction I find it necessary to consult other sources. That should not be the case for a wiki article. Would someone please write an introductory sentence or two that would explain in lay terms what a multiple correlation is. For instance, for a simple correlation coefficient it could be said that it is "a measure of the degree of relatedness of two variables. A coefficient of 1 indicates that they are mutually dependent, so that for one variable to change the other must also, whereas a coefficient of zero denotes complete independence.". So ... what is multiple correlation? Is it the correlation of multiple variables and if so is it some kind of average or ...? LookingGlass (talk) 14:14, 18 October 2015 (UTC)


 * Here's a starter for anyone up to it http://onlinestatbook.com/2/regression/multiple_regression.html (p.s a reader doesn't want to take a course; just to have the bare bones of it) LookingGlass (talk) 14:22, 18 October 2015 (UTC)


 * Hope my changes today help some. If not, please make specific suggestions for further improvement. Thanks. Loraof (talk) 16:55, 6 November 2015 (UTC)