0
$\begingroup$

I have theoretical knowledge of why we should use bonferroni; however, unfortunately not how.

I am conducting multiple t.tests, sometimes within the main data (number of t.tests = 5, number of main tests/regression = 2), and sometimes within the split data by gender (number of tests = 2).

I am wondering then whether I should conduct bonferroni separately for the main data and the split data.

using R, finding the adjusted p using bonferroni for the main data is then:

p.adjust(c(0.0012, 0.005, 0.0464, 0.003, 0.0321, 0.001, 0.002), method = "bonferroni")

and the same for the split data would then be:

p.adjust(c(0.0464, 0.0321), method = "bonferroni")

Am I doing this correctly? Any help will be appreciated, thanks.

$\endgroup$
  • $\begingroup$ And am I understanding it correctly that it is generally not preferred to do bonferroni on exploratory analyses? Thanks! $\endgroup$ – user240313 May 29 '19 at 18:03
0
$\begingroup$

The Bonferroni correction is an adjustment to the threshold of significance (i.e., alpha) for your p-value. This correction is made to account for inflated Type I error (the higher the chance for a false positive). It is a very straight forward statistic -- you simply divide your p-value by the number of independent hypothesis tests that you have conducted. If you are testing 9 hypotheses, and the standard alpha value in your field is .05, your Bonferroni corrected alpa is .05/9 = .006. This means any p-value you have that is above .006 (but below .05) is treated with suspicion.

Thus, your first step is to really specify how many hypotheses you are testing.

I hope that helps!

$\endgroup$
  • $\begingroup$ Yes that helps enormously, thanks!! $\endgroup$ – user240313 May 29 '19 at 18:47

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.