Skip to content
Prev 363105 / 398502 Next

p.adjust not working correctly?

Part of the confusion is that Bonferroni is often described as an adjustment to the significance level (sig-level) not the p-value. For example, to evaluate p-values when there are 8 tests, we compare the p-value to the sig-level/8 so the sig-level decreases. The p.adjust() function adjusts the p-value instead of the sig-level so we multiply the p-value by the number of tests (p-value * 8). As a result the p-values increase.

-------------------------------------
David L Carlson
Department of Anthropology
Texas A&M University
College Station, TX 77840-4352

-----Original Message-----
From: R-help [mailto:r-help-bounces at r-project.org] On Behalf Of S Ellison
Sent: Wednesday, August 3, 2016 6:05 AM
To: Alicia Ellis; r-help at r-project.org
Subject: Re: [R] p.adjust not working correctly?
You should re-check what a Bonferroni correction does, or at least reboot your intuition (mine needs rebooting all the time). All the p-values should _increase_ by a factor of n, with a ceiling of 1.0. Dividing by n would imply incorrectly that individual events have become less probable as the number increases.

The result you have obtained is what is supposed to happen.

S Ellison
*******************************************************************
This email and any attachments are confidential. Any use...{{dropped:9}}