How do I interpret FDR values from mafdr?

51 views (last 30 days)
Amy Frithsen
Amy Frithsen on 13 Sep 2016
Answered: Mina Kheirkhah on 6 May 2020
I have a list of p-values that came from ttest2. I've attached the list here (called 'uncorrected_p_values').
When I use the following command: FDR = mafdr(uncorrected_p_values) I get the values listed in the attached file called 'FDR'.
How do I interpret these values? I thought these were just the adjusted p-values (such that if the FDR value is below .05, then the result is significant), but this can't be correct. First off, the FDR values are way higher than I would expect - an example is that an uncorrected p value of .065 gets converted to an FDR value of .9556. Also, I read somewhere that the default option of setting 'BHFDR' to false is not very wise with a list of p values less than ~1000. I only have 120 p values. Even setting BHFDR to true, I still get really high values for FDR. Do I need to be looking at the q values, and if so, how are these interpreted? How do I know which original p values I should consider significant?
Sorry for the naive questions, I just really have no idea what the output of this function means even after reading through the help section. Thanks for any advice!
~ Amy

Answers (1)

Mina Kheirkhah
Mina Kheirkhah on 6 May 2020
Hello Amy,
I had the same problem as you and then I decided to use another kind of FDR (Benjamini & Hochberg false discovery rate) which has more explanation in MATLAB website and it is easier to understand:
Hope it helps,


Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by