Talk:Family-wise error rate
![]() | This article is rated Start-class on Wikipedia's content assessment scale. It is of interest to the following WikiProjects: | ||||||||||||||||||||
|
Merger proposal
editThe merge template from April 2009 did not point to a valid place for discussion: now fixed. I did not find any discussion apart from a note that the two things seemed the same. Melcombe (talk) 11:57, 20 July 2010 (UTC)
The two things are the same, it is strictly a synonym and the page should include both names and a complete description of both. Korymath (talk) 10:05, 13 Septebmer 2011 (UTC)
making FWER definition more accurate
editIn this article, FWER is defined as
FWER = Pr(V ≥ 1)
However, as usual in hypothesis testing, the probability distribution which is used for computing what is denoted "Pr" extremely matters. Thus in a way which is consistent with the type I risk, the definition of FWER cannot avoid mentionning that "Pr" is computed using a probability distribution consistent with all null hypotheses which are assumed to be true. Moreover, there may be more than one such probability distribution and FWER may not be uniquely defined by the above equality. In that case, which of these should we pick up for computing Pr(V ≥ 1) ? --81.66.237.175 (talk) 14:11, 15 February 2012 (UTC)
I have just finished elaborating the article
editHi everyone, as part of my statistics master's "selective inference" coursework I was asked to elaborate this article. First, I clarified the definition part and added the necessary distinction between weak and strong FWER control that is mentioned above. I have written several new sections, such as the "controlling procedures" section, and also separated between the example that was shown before under "what Constitutes a family" and a new part called "the concept of a family". Regarding the example - I've omitted complex wording and also made it a little more readable with bullets, spacing and such.
Miron.Avidan (talk) 00:17, 03 August 2012 (CET)
significant VS non-significant?
editI'm not sure about the concept of "declaring" a hypothesis significant or non-significant. significant or non-significant is rather the value of the statistic used to test the hypothesis. I would therefore say null hypothesis "rejected" in place of "declared significant" and null hypothesis "accepted" in place of "declared non-significant". 129.240.215.211 (talk) 09:34, 20 December 2012 (UTC)
Link to Pairwise Comparisons
editI'm unsure where to post this, so feel free to put it elsewhere. When I click on "pairwise comparisons" it leads me to http://en.wikipedia.org/wiki/Pairwise_comparison, where it says "This article is about pairwise comparisons in psychology. For statistical analysis of paired comparisons, see paired difference test."
Shouldn't the link lead to "paired difference test" instead? — Preceding unsigned comment added by 132.252.228.226 (talk) 13:25, 24 October 2014 (UTC)
- Hi there. The link is o.k. Paired difference test is a specific statistical test. Pairwise comparison is the issue where you are comparing many different groups to one another. That article probably should not be just about psychology, but rather on that general issue (for example, it may be also used when comparing drugs).
- Cheers, Tal Galili (talk) 14:56, 24 October 2014 (UTC)
Difference Holm vs. Hochberg
editThe way they are presented here, how are Holm's step-down and Hochberg's step-up procedure different, except for moving terms in the denominator around? — Preceding unsigned comment added by 2001:16B8:2E3A:4500:B0C0:568B:4CAC:3D2 (talk) 15:47, 10 March 2022 (UTC)