threshold by p or q, instead of t/F value by 3dclust/3dmerge

AFNI Message Board
1

Hi afni group:

Greetings!

I wander if there is one way to threshold a statistical dataset by setting the q/p value, instead of using t/F value, by 3dclust/3dmerge, or you have other recommendation.
The reason I am looking for this method is that for different contrast, the statistical threshold could change with the same q value.
Many thanks!

Mengxing

2

Hi, Mengxing-

Well, you can use 3dClusterize to threshold more easily than with 3dmerge/3dclust, probably:
https://afni.nimh.nih.gov/pub/dist/doc/htmldoc/programs/3dClusterize_sphx.html#ahelp-3dclusterize
There, you can use either voxelwise p or statistic value, along with cluster volume. For a code example using it on an actual dataset with 3dClustSim and , you can see here:
https://afni.nimh.nih.gov/pub/dist/doc/htmldoc/codex/main_det_2018_ChenEtal.html

I haven’t looked at using this specifically with 3dFDR output or whatever you are using, but it should be an easier place to start than 3dclust/3dmerge.

–pt

3

Hi pt,

your reply is super helpful, while I still wander, if there is one way to threshold by specifying the q FDR value?

best,
Mengxing

4

Hmm, I’m not certain, but I think you can use “fdrval -qinput …” to get the corresponding statistic threshold out of a specific brick when you provide a qvalue of interest?
https://afni.nimh.nih.gov/pub/dist/doc/program_help/fdrval.html

–pt

5

Perfect, that will do by assigning the std to a variable and use it in 3dclust!!!
You work at weekends! Take a rest! although I am sooooo happy to have your reply without waiting for Monday.

best,
Mengxing

6

I had a bit of spare time between World Cup matches…

–pt