Question about fdr q value

AFNI Message Board
1

Hi,

If I understand correctly, q value is calculated by number of false positives / number of voxels that pass threshold. My question why is my q value stops changing at some point when I keep decrease p value, so say from p value is 0.1 to 0.0001, q value is always 0.2. Does it mean the number of FPs and number of voxels that pass threshold change proportionally? How is this possible?

Thanks!

2

Hello,

So at p=0.1, one would expect 10 out of every 100 voxels to survive threshold. In your case there were 50 out of 100, giving q = 0.2 (10 expected/50 surviving). But maybe at p=0.01, when expecting 1 out of every 100 voxels, you only had 4. That would give q=0.25. But q is not allowed to increase, so it is still 0.2.

And maybe the rate just never got better, because there are not enough super-threshold voxels.

See these other posts:

[Re: FDR correction on 3dttest output](https://afni.nimh.nih.gov/afni/community/board/read.php?1,73957,73986)
[Re: small p-value / large q-values](https://afni.nimh.nih.gov/afni/community/board/read.php?1,68991,68996)
  • rick
3

Oh OK, I missed the q is not allowed to increase part. Thanks!