gravatar for ATpoint

1 hour ago by

What you see is expected. The Bonferroni correction works by multiplying the pvalues with the number of tests, here 3294.

#/ Example (four values, so everything gets multiplied by 4, and 1 stays 1 of course):
p.adjust(c(0.05, 0.0005, 0.1, 1), "bonferroni")
> 0.200 0.002 0.400 1.000

That means that in order to get an adjusted p-value of 0.05 or smaller your uncorrected p-values must be at least 0.05/3294 = 0.00001517911. All larger p-values will turn non-significant if using 0.05 as cutoff. It can well be that your experiment does not have the statistical power to return these kinds of significances. Bonferroni is quite a stringent correction method. You might want to look into the Benjamini-Hochberg method, but if you lack statistical background knowledge it is generally best to either work with a statistician who gives counsel or to use software packages that do the stats work for you internally.

What kind of experiment is that you are analyzing?


modified 1 hour ago

1 hour ago


Source link