Talk:One- and two-tailed tests

Please clarify and provide reference[edit]

A two-tailed test is used when the two samples being compared are related in some way. For example, you would use a two-tailed test if one random sample was 15 quarter horses and the second sample was 15 sires or dams of those same horses. A one-tailed test is appropriate if no known relationship exists between the samples, for example, two random samples of 15 unrelated quarter horses. --206.208.110.32 20:58, 17 August 2005 (UTC)[reply]

Is this actually correct? It doesn't seem to agree with the formal definitions I could find. --Jethero 15:48, 1 December 2005 (UTC)[reply]

I don't think so. I have to admit that I don't know much about this, but everything I've seen, and all the people I've talked to that actually do know about this disagree with 206.208.110.32's definition. 66.245.2.167 02:37, 31 January 2006 (UTC)[reply]

The definition by 206.208.110.32 above is actually talking about "paired" samples and has nothing to do with "two-tailed" tests.

I'm doing some research using statistical analysis, and the comment immediately above is correct. Tests for samples with a known relationship are "paired t-tests," while tests without a relationship are "unpaired." One-tailed tests (for either paired or unpaired variants) look for data that is too unlike the central tendency of the distribution, but in only one direction. Two-tailed tests look for data that is too unlike the central tendency, either too high or too low. For example, manufacturers may do an analysis of products that fall beneath a certain quality criterion, but they probably do not worry about items of higher than average quality. This would involve a one-tailed t-test.Alhead (talk) 21:34, 11 July 2008 (UTC)[reply]

This seems an reasonably common confusion, so I’ve added a link to Paired difference test in the “See also” section as of this edit. Hope this helps!

—Nils von Barth (nbarth) (talk) 21:27, 23 March 2013 (UTC)[reply]

Renaming this page[edit]

Problem: if you type in "one-tailed" it just says there is no such page. It would not make sense to create a separate page to define "one-tailed"; both one-tailed and two-tailed are more appropriately discussed on the same page. But, it wouldn't work very well to redirect "one-tailed" to this page, because someone might get confused and think "one-tailed" means the same thing as "two-tailed". I suggest renaming this page to "One-tailed and two-tailed tests" and putting redirects so anyone typing in "two-tailed", "two-tailed test", "two tailed test", "one-tailed", etc. etc. will find this page. What do others think -- is it OK if I rename it? --Coppertwig 12:51, 21 November 2006 (UTC)[reply]

I agree --Jdedmond 12:09, 24 March 2007 (UTC) Also, what about a P value table/ example of a type of 2 tailed test (Mann-Whitney U)?[reply]

Done.Fgnievinski (talk) 19:29, 20 December 2011 (UTC)[reply]

On tailed and two tailed Ttest[edit]

I'm new to this site. but have a question. I have to complete a Ttest and do not know how to determine if it is one tailed or two tailed and the type. Can anyone help me?

Robyn7 (talk) 04:07, 29 April 2008 (UTC)[reply]

Perhaps my comment above can help. Alhead (talk) 21:36, 11 July 2008 (UTC)[reply]

Robyn7: you can easily identify these tests by their specific values. — Preceding unsigned comment added by 139.190.191.94 (talk) 18:59, 29 May 2012 (UTC)[reply]

Coin flipping example[edit]

This example is very confusing to me. Could someone tell me, what the p-value would be in the one-tailed and the two-tailed cases, respectively, if my sample was TTTTT? Timur lenk (talk) 20:21, 27 October 2021 (UTC)[reply]