Talk:Interaction (statistics)

A graph should have been displayed here but graphs are temporarily disabled. Until they are enabled again, visit the interactive graph at pageviews.wmcloud.org

Michael Hardy: I do not understand why you replace the word residuals (neutral) with the word errors then qualify the word errors as being potentially misleading! Residuals just means the difference between the model and the observations. What's your problem? Cutler 01:39, 8 Feb 2004 (UTC)

Because that's not what "residual" means. As I said (and you act as if you did not read it) a residual is not the same thing as an error: a residual is an observable estimate of an unobservable error. Why didn't you read that before asking me why I changed it. Michael Hardy 21:50, 8 Feb 2004 (UTC)

OK, now I've created a new article, errors and residuals in statistics, which corrects Cutler's mistake. Michael Hardy 02:34, 9 Feb 2004 (UTC)

Good article. Apologies that I didn't articulate my point as sharply as I chose my wording. Cutler 19:26, 9 Feb 2004 (UTC)

Proposal for merger[edit]

Greetings all! I propose that the content in the 'Statistics' section of the Interaction page be merged with the content on this page,and that this page be the destination for the statistics component of the forthcoming 'Interaction' disambiguation page. Please post any comments/ideas/concerns that you might have regarding this proposal. --DanSoper 08:33, 12 June 2006 (UTC)[reply]

Proposal for link[edit]

Hello everyone,

I propose that this page be linked to the term "effect modification", which is the term epidemiologists use when referring to statistical interaction. I would like to be able to type effect modification in the search field, and be sent to this wiki, except that I don't know how to create that linkage.

James

I've created a redirect from effect modification to this article. Michael Hardy 21:17, 28 November 2006 (UTC)[reply]

In the Rothman epidemiology book he states that interaction has a different meanings in statistics and epidemiology. In epidemiology he calls it effect-measure modification. Can anobody ad an explanation about this to the article? Just linking doesn't explain anything about different uses. Pizzaman79 (talk) 08:13, 24 April 2009 (UTC)[reply]

Huh?[edit]

I posted the following at talk:interaction variable:

The second sentence in this article says an interaction variable is formed by multiplying two predictor variables.

But then later in the article it's talking about categorical predictors, which clearly cannot be multiplied.

Now someone's proposed merging this with interaction (statistics). And that article also begins by talking about multiplication of two predictors, just as if it's about that particular form of interaction rather than about interaction in general.

What a mess.

Michael Hardy (talk) 01:29, 13 May 2009 (UTC)[reply]

Not only that, but Interaction variable contains a serious misunderstanding of p-values; it claims "the ANOVA-RM will produce values indicating the probability that the difference is due to there being an actual difference (as opposed to the difference being the result of coincidence)". Yikes. Btyner (talk) 02:04, 6 June 2009 (UTC)[reply]

Discussion of merge - May 2009[edit]

Those thinking of merging this with interaction variable should probably think also about Moderation (statistics), Moderator variable and Mediation (statistics) some of which overalps here but starting from a "psychometric" point of view. Melcombe (talk) 16:08, 13 May 2009 (UTC)[reply]

I went ahead and merged Moderator variable into Moderation (statistics), so that's one fewer source of overlap with which to contend. Btyner (talk) 00:09, 15 May 2009 (UTC)[reply]

ANOVA example accuracy?[edit]

The description of implications of the first table seems not to apply the information in the table. I don't know the concept (hence why I'm reading the article), but it seems that A in the example in every case brings down the average value; can it still be additive, as it is described? —Preceding unsigned comment added by 67.193.142.111 (talk) 00:17, 12 February 2010 (UTC)[reply]

The additivity is reflected in the fact that moving A from 1 to 0 has the same effect when B=1 as when B=0 -- in both cases, the mean response increases by 2 units. This is a necessary and sufficient condition for additivity to hold. Skbkekas (talk) 19:52, 12 February 2010 (UTC)[reply]

this is not consistent with what is explained below the first table... very confusing! —Preceding unsigned comment added by 157.193.1.210 (talk) 14:44, 1 October 2010 (UTC)[reply]