2
$\begingroup$

I am performing a RQL analysis ("http://www.esapubs.org/archive/ecol/E095/002/suppl-1.pdf"). This analysis meassures the relationship between different traits of three different matrix showing if there is a positive or negative relationship. This RQL analysis make a fourth-corner method.

Well, this method have the option to make a p-adjust method to solve multiple comparison problem. However, my variables are independent and I belive that I should not apply to my dataset. Am I right? Or always that two or more matrix are compared I should apply a multiple comparison adjustment?

Thanks in advance.

$\endgroup$
3
$\begingroup$

The p-value adjustments corrects the inflated number of significant results when performing many significance tests.
If you have a test with 5 % false positive rate and but you run it several times the chances of getting a false positive results in any of these tests is far greater than 5 %. For example repeating the test with 5 % false positive rate 20 times means that there is a 60 % chance of at least one false positive result.

You can see this in the tutorial you have linked to in your question, where the number of significant differences goes down from 26 without correction to 18 with correction.

Therefore, if you are performing multiple comparisons you should do a correction. The problem of inflated p-values will occur when you are performing many tests even if the tested variables are independent.

$\endgroup$
0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.