My Math Forum

My Math Forum (http://mymathforum.com/math-forums.php)
-   Advanced Statistics (http://mymathforum.com/advanced-statistics/)
-   -   Tukey test (http://mymathforum.com/advanced-statistics/342919-tukey-test.html)

shaharhada November 24th, 2017 07:05 AM

Tukey test
 
I read the definition in Wikipedia and I don't understand a thing.
What is the Tukey test in a few words in Simple English!!!
Thank for answering or considering my post.

James White November 29th, 2017 10:41 AM

When you compare multiple groups (such as group 1, group 2 and group 3), ANOVA is used. If ANOVA produces p<0.05, you would further like to know which of the following three pairs have 'honestly' significant differences:
  • group 1 vs group 2;
  • group 2 vs group 3;
  • group 1 vs group 3.

If you use t-test, there will be more false positives than expected as 1 ANOVA test now becomes 3 t-tests. For example, if alpha=0.05 is set for 1 ANOVA test, alpha will become 1-0.95*0.95*0.95=0.14 for 3 t-tests. This is commonly known as 'alpha inflation' in multiple comparisons.

To avoid 'alpha inflation' in multiple comparisons, Tukey's test is used to keep the alpha at the same level and therefore the results are "Honestly Significant Differences (HSD)". This online Tukey's HSD calculator gives explanations for some real data examples.

shaharhada November 30th, 2017 06:27 AM

Why on the alpha level need to be keeped?
Thanks on your answer!!!

James White December 1st, 2017 03:39 PM

Quote:

Originally Posted by shaharhada (Post 585058)
Why on the alpha level need to be keeped?
Thanks on your answer!!!

It is preferred to keep alpha at the same level (usually set at 0.05). Otherwise, people can get 'significant results' by simply making a large amount of comparisons.

This famous research study gives you a simple explanation:
Why Most Published Research Findings Are False


All times are GMT -8. The time now is 01:53 PM.

Copyright © 2019 My Math Forum. All rights reserved.