Talk:Correlation ratio

[This article is] poorly explained. --Eequor 03:39, 22 Aug 2004 (UTC)

Cleanup request
This article is incomprehensible to anyone unfamiliar with advanced statistical notation. It could be repaired with a concrete example and an intuitive, rather than symbolic or mathematical explanation. A graph or two would also help a lot. -- Beland 01:54, 23 August 2005 (UTC)


 * This is really part of a wider problem: currently large areas of the topic of statistics are very deficiently treated on Wikipedia. A few statistics articles are really good, some others are quite competent, but many are skimpy stubs, and this is one of those.  I may return here. Michael Hardy 02:27, 25 August 2005 (UTC)

Please include more information


 * I have attempted an example --Rumping 00:35, 30 September 2007 (UTC)

Anyone understands this thoroughly?
It is worth noting that if the relationship between values of $$x \;\ $$ and values of $$\overline{y}_x$$ is linear (which is certainly true when there are only two possibilities for x'') this will give the same result as the square of the correlation coefficient, otherwise the correlation ratio will be larger in magnitude. It can therefore be used for judging non-linear relationships.''

If you understand this thoroughly, please help improving this article and put it more clearly. I am very unhappy with this sentence, since I am not understanding it. I do not want to remove it, since there needs to be a reference made to the correlation coefficient and it seems that the original author knew about it. It would just be nice to convey this message comprehensively. Tomeasy (talk) 11:32, 29 March 2008 (UTC)

Standard deviation
I have taken the later of the two following statements out, since I think it is wrong.
 * $$\eta^2 = \frac{\sum_x n_x (\overline{y}_x-\overline{y})^2}{\sum_{xi} (y_{xi}-\overline{y})^2}$$

which might be written as
 * $$\frac{{\sigma_{\overline{y}}}^2}{{\sigma_{y}}^2}.$$

The use of the sigma's implies that we should be dealing with standard deviations here. However, substituting their definition does not yield the given equation. Also it is not very clear to me, what standard deviation is precisely meant in the numerator. I tried interpreting it as the standard deviation of the category means, but it failed to yield the stated equation. Correct me, if I am wrong. Tomeasy (talk) 11:41, 29 March 2008 (UTC)


 * It doesn't look so bad to me, though there might be a need for further description. Clearly
 * $${\sigma_{y}}^2 = \frac{\sum_{xi} (y_{xi}-\overline{y})^2}{n}$$.
 * But $${\sigma_{\overline{y}}}^2$$ needs a definition. I would guess it was intended to be a weighted variance, using $$n_x$$ as the weights, of the form
 * $${\sigma_{\overline{y}}}^2 = \frac{\sum_{x} n_x (\overline{y}_x-\overline{y})^2}{\sum_{x} n_x} $$.
 * Then division gives the result --Rumping (talk) 22:29, 2 August 2008 (UTC)


 * Looks convincing to me. So, if you want to include it with the appropriate definition of $${\sigma_{\overline{y}}}^2$$, please go ahead. T om ea s y T C 08:35, 3 August 2008 (UTC)

Merger proposal
I have proposed that this article be merged into intraclass correlation coefficient. The correlation ratio statistic and its population analogue are the same as the ICC (more specifically, the statistic is one of several ICC's in use, and the population value is identical). This particular use of the ICC is different from the usual uses of the ICC, so I would propose adding a new section to the ICC page explaining how the ICC can be used to identify non-linear relationships when data are observed with replication, and stating that the term "correlation ratio" is sometimes used when the ICC is applied in this way. Skbkekas (talk) 01:35, 15 June 2009 (UTC)

Links to other languages
I've tried unsuccessfully to add a link to the same page in Spanish. Could someone do it? thanks — Preceding unsigned comment added by 190.193.138.41 (talk) 00:09, 28 May 2013 (UTC)