Having 10 behavioral variables and 5 independent parameters of brain activation, I've calculated linear correlations between each variable and parameter, obtaining a 10x5 matrix of r-values, and a 10x5 matrix of p-values.

I want to know which correlations are significant after correction for multiple comparisons.


The correct way to correct for multiple comparison, using the matrix of p-values, is:

  1. By row (10 separate corrections), or
  2. By column (5 separate corrections), or
  3. On the whole matrix (1 correction)?
  • $\begingroup$ There are many ways how you can do it. Here are two: If you are using R you can use the corr.test() function in the psych package (see here); or you can adjust manually by using a sequential Bonferroni adjustment for example (see here). $\endgroup$ – Stefan Dec 8 '16 at 18:46
  • 1
    $\begingroup$ I know how to do that. I was asking which is the correct way, between the three proposed approaches. $\endgroup$ – smndpln Dec 8 '16 at 19:10
  • 1
    $\begingroup$ I think that must depend on what the underlying science is. If the brain activation is for separate regions which have distinct functions then it seems hard to justify adjusting them for one another (for instance). $\endgroup$ – mdewey Dec 11 '16 at 13:45
  • $\begingroup$ The parameters of brain activations, as I wrote, are indipendent, and possibily linked to different brain functions. Do you think this is a valid reason to choose approach 2? (that means, 5 separate corrections, one for each roi) $\endgroup$ – smndpln Dec 11 '16 at 14:28

Your Answer

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

Browse other questions tagged or ask your own question.