New subject: Question on GLM - correlation

3 Dec 2019


      That does not have the pvr in it. You can extract the pvr for that vertex and add it as a column to the X matrix with
pvr = fast_vol2mat(MRIread('pvr.mgz'));
pvrvtx = pvr(:,vertexno+1);
ps. Please remember to post to the list
On 12/3/2019 1:29 PM, Barletta, Valeria wrote:
Ok done.  This is the Xg.mat
1    29
     1    44
     1    38
     1    47
     1    34
     1    25
     1    30
     1    22
     1    52
     1    31
     1    37
     1    42
     1    28
     1    39
     1    44
     1    44
     1    32
     1    32
     1    36
     1    47
     1    52
And the cond(Xg'*Xg) is again:   3.0555e+04
Looks like there is no extra column for the pvr in the Xg.mat
________________________________
From: Greve, Douglas N.,Ph.D. DGREVE@mgh.harvard.edumailto:DGREVE@mgh.harvard.edu
Sent: Tuesday, December 3, 2019 11:03 AM
To: Barletta, Valeria VBARLETTA@mgh.harvard.edumailto:VBARLETTA@mgh.harvard.edu; Freesurfer support list freesurfer@nmr.mgh.harvard.edumailto:freesurfer@nmr.mgh.harvard.edu
Subject: Re: [Freesurfer] Question on GLM - correlation
I'm not sure those condition numbers include the pvr. Try loading the Xg.mat into matlab and computing
cond(Xg'*Xg)
Make sure Xg has the extra column for the pvr
On 12/3/2019 10:42 AM, Barletta, Valeria wrote:
Ok it worked.
So I tried with a non-zero vertex and I got:
Normalized matrix condition is 81.1382
Matrix condition is 30558.5
Found 149926 points in label.
Found 149926 voxels in mask
Saving mask to /autofs/space/lipari_001/users/valeria/NODDI/GLM//GLM_CME_ficvf_doug/rh_contrast-MS_CME_all_FIS_ICA1/mask.mgh
Reshaping mriglm->mask...
search space = 74490.928733
DOF = 18
Dumping voxel 1024 0 0 to /autofs/space/lipari_001/users/valeria/NODDI/GLM//GLM_CME_ficvf_doug/rh_contrast-MS_CME_all_FIS_ICA1/voxdump-1024-0-0
...and with a zero vertex and got:
Normalized matrix condition is 81.1382
Matrix condition is 30558.5
Found 149926 points in label.
Found 149926 voxels in mask
Saving mask to /autofs/space/lipari_001/users/valeria/NODDI/GLM//GLM_CME_ficvf_doug2/rh_contrast-MS_CME_all_FIS_ICA1/mask.mgh
Reshaping mriglm->mask...
search space = 74490.928733
DOF = 18
Dumping voxel 150252 0 0 to /autofs/space/lipari_001/users/valeria/NODDI/GLM//GLM_CME_ficvf_doug2/rh_contrast-MS_CME_all_FIS_ICA1/voxdump-150252-0-0
________________________________
From: freesurfer-bounces@nmr.mgh.harvard.edumailto:freesurfer-bounces@nmr.mgh.harvard.edu freesurfer-bounces@nmr.mgh.harvard.edumailto:freesurfer-bounces@nmr.mgh.harvard.edu on behalf of Greve, Douglas N.,Ph.D. DGREVE@mgh.harvard.edumailto:DGREVE@mgh.harvard.edu
Sent: Monday, December 2, 2019 2:56 PM
To: freesurfer@nmr.mgh.harvard.edumailto:freesurfer@nmr.mgh.harvard.edu freesurfer@nmr.mgh.harvard.edumailto:freesurfer@nmr.mgh.harvard.edu
Subject: Re: [Freesurfer] Question on GLM - correlation
oh, sorry, did not see the pvr. It could very well be the pvr that is causing the patchiness. Unfortunately, you have to debug it vertex by vertex. I would choose a few vertices (one with signal, on without) and run mri_glmfit with --voxdump vertexno 0 0 This will create a subfolder with the GLM for that vertex. I'm not sure what to tell you to look for, but I would at least check the condition numbers in each vertex. What is the nature of ficvf?
On 12/2/2019 1:22 PM, Barletta, Valeria wrote:
My other continuous variable is --pvr ${hemi}.ficvf.mgh
Not sure about --no-prune, I found it in my script as a default... Shall I remove it?
________________________________
From: freesurfer-bounces@nmr.mgh.harvard.edumailto:freesurfer-bounces@nmr.mgh.harvard.edu freesurfer-bounces@nmr.mgh.harvard.edumailto:freesurfer-bounces@nmr.mgh.harvard.edu on behalf of Barletta, Valeria VBARLETTA@mgh.harvard.edumailto:VBARLETTA@mgh.harvard.edu
Sent: Friday, November 29, 2019 3:08 PM
To: Freesurfer support list freesurfer@nmr.mgh.harvard.edumailto:freesurfer@nmr.mgh.harvard.edu
Subject: [Freesurfer] Question on GLM - correlation
Dear Freesurfers,
I am running glmfit to assess for correlation between y and another continue variable across the cortex (ficvf), correcting for the effect of age, using the command:
mri_glmfit \
  --y {surfname_cme}.mgh \
  --fsgd {contrast}.fsgd \
  --C {contrast}.mat \
  --pvr ${hemi}.ficvf.mgh" \
  --surf fsaverage $hemi \
  --glmdir ${glm_dir}/${hemi}_contrast-${contrast}_${surfname_cme} \
  --no-prune
This is my matrix:
1 0 0
And my fsgd file:
Title CME
Class MS
Variable age
Input MS069_R01_01  MS  29
Input MS070_R01_02_R1year  MS  44
Input MS072_R01_04  MS  38
Input MS073_R01_05 MS 47
Input MS075_R01_07_R1year MS 34
Input MS076_R01_08 MS 25
Input MS077_R01_09 MS 30
Input MS078_R01_10 MS 22
Input MS080_R01_12 MS 52
Input MS081_R01_13 MS 31
Input MS082_R01_14 MS 37
Input MS084_R01_16 MS 42
Input MS086_R01_18 MS 28
Input MS087_R01_19_R1year MS 39
Input MS088_R01_20 MS 44
Input MS089_R01_29 MS 44
Input MS090_R01_21 MS 32
Input MS091_R01_22 MS 32
Input MS092_R01_23 MS 36
Input MS094_R01_25 MS 47
Input MS104_R01_36 MS 52
This is the result:
[cid:part10.AAFB4919.2AE170D2@mgh.harvard.edu][cid:part11.F1A0EFEC.57C56985@mgh.harvard.edu]
Am I proceeding correctly?
Thank you,
Valeria
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edumailto:Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer

Re: [Freesurfer] Question on GLM - correlation

Thanks, sorry for the long interaction...

Thanks, sorry for the long interaction...

Thanks, sorry for the long interaction...