corrections in FIVE vs peak info

60 views
Skip to first unread message

Harry Moore

unread,
Oct 22, 2019, 4:20:42 AM10/22/19
to Aaron's fMRI matlab tools.
Hello,

I've noticed that when visualising some overlays in FIVE that sometimes when I apply an FWE or FDR correction FIVE corrects the uncorrected p value, and also alters the activated voxels in the overlay, but when I either select Options > Get Peak Info, or run peak_nii, the 'sigthresh' variable does not contain any values and therefore leads me to conclude that there are actually no voxels that survive the FWE or FDR correction - so am I right in trusting the sigthresh result rather than the uncorrected p value when I press FWE/FDR in the FIVE viewer?

Thanks

Aaron Schultz

unread,
Oct 22, 2019, 11:36:51 AM10/22/19
to fmri_mat...@googlegroups.com
I will leave it to Donald to answer definitively, but the main difference will have to do with voxel wise corrections vs cluster wise corrections. 

-Aaron

--
You received this message because you are subscribed to the Google Groups "Aaron's fMRI matlab tools." group.
To unsubscribe from this group and stop receiving emails from it, send an email to fmri_matlab_to...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/fmri_matlab_tools/1c1591fb-4200-4ffb-9a07-08ffdb5f50c9%40googlegroups.com.
Message has been deleted
Message has been deleted

H M

unread,
Nov 4, 2019, 12:09:59 PM11/4/19
to Aaron's fMRI matlab tools.
Hi again Aaron,

I have been looking at the 'voxelstats' variable created by peak_nii and 'get peak info' in FIVE to get around this problem, and I understand that when I apply an FDR correction in FIVE/peak_nii it calculates the uncorrected p value necessary to meet the FDR threshold (in my case 0.05). But when looking at the output variable 'voxelstats', although the uncorrected p value column matches up with what I expected the uncorrected p values to be at an FDR threshold of 0.05, I expected the q value column to only have values of 0.05 and lower, which is not the case. It's no unlikely that I've misunderstood the theory behind the correction, but do you think that makes sense? (for example,  the uncorrected p value for FDR = 0.05 was set to 0.002, which was the highest value in the uncorrected p value column in 'voxelstats' and its corresponding q value was 0.49)

Cheers


On Tuesday, 22 October 2019 16:36:51 UTC+1, Aaron Schultz wrote:
I will leave it to Donald to answer definitively, but the main difference will have to do with voxel wise corrections vs cluster wise corrections. 

-Aaron
On Tue, Oct 22, 2019 at 4:20 AM Harry Moore <htam...@gmail.com> wrote:
Hello,

I've noticed that when visualising some overlays in FIVE that sometimes when I apply an FWE or FDR correction FIVE corrects the uncorrected p value, and also alters the activated voxels in the overlay, but when I either select Options > Get Peak Info, or run peak_nii, the 'sigthresh' variable does not contain any values and therefore leads me to conclude that there are actually no voxels that survive the FWE or FDR correction - so am I right in trusting the sigthresh result rather than the uncorrected p value when I press FWE/FDR in the FIVE viewer?

Thanks

--
You received this message because you are subscribed to the Google Groups "Aaron's fMRI matlab tools." group.
To unsubscribe from this group and stop receiving emails from it, send an email to fmri_mat...@googlegroups.com.
Reply all
Reply to author
Forward
0 new messages