Talk:Moderation (statistics)

From Wikipedia, the free encyclopedia
Jump to: navigation, search

Untitled[edit]

I am not sure how to do this but. Somebody should ask that this article be merged with moderator variable. In either case, both are also stubs and should be reported that way on the top. I would say the importance should also be changed to at least Medium if not high. Moderation is a key principle behind describing relationships in Psychology. I would expand it myself but I also get moderation/mediation confused. —Preceding unsigned comment added by Nonymous-raz (talkcontribs) 16:34, 11 May 2009 (UTC)

Done. Could still use some cleanup. Btyner (talk) 00:00, 15 May 2009 (UTC)

Multicollinearity in Moderated Regression: in this part, "IV" & "DV" are mentioned, but it is never explained, what these stand for. Nikopoloopkin (talk) 09:55, 12 June 2012 (UTC)

Fixed. Duoduoduo (talk) 22:23, 29 January 2013 (UTC)

Subsection on Centering and standardizing[edit]

I'm going to delete the subsection "Centering and standardizing" of the section "Multicollinearity in moderated regression". It falsely claims that we can avoid the problems caused by multicollinearity simply by linearly transforming the individual variables. But even in the presence of interaction terms, the revised regression is just a linear transformation of the original regression, and all results for parameters are just linear transformations as well. Thus one version's results can be exactly derived from the other's. And the standard errors of all the estimated effects are identical. So while it's true that the transformed version has less multicollinearity, the resulting favorable effect on the standard errors is exactly offset by the unfavorable effect due to smaller variance of the interaction variable. Duoduoduo (talk) 22:23, 29 January 2013 (UTC)