Dear FS experts,
I finished running the GLM and I wonder what is the best way to further analyse my data in order to see if there is any relation between age (at seizure onset) and cortical thickness for my patients versus control group.
1). Why when I overlay the sig.mgh in tksurfer lh inflated, I get more clusters than the number of clusters obtained from mri_glmfit-sim?
2). What sig.mgh represents?
3) What's the real significance of a cluster (how a cluster is formed)?
4). Why my thickness value in clusters (eg cluster no. 1 which coresponds to the posteriorcingulate) for subject no.1 is different that the thickness value obtained for the same region in the lh.aparc.stats while running recon-all?
5). After running the GLM should I use visualizing and plotting method to further analyse my data and load FSGD file lh.gender_age.glmdir/y.fsgd?
6). Should my ROI's be defined or be the same with my clusters?
7). What is the difference between GLM and Qdec? What method is the best to analyse the relation between age and cortical thickness for my patients versus control group?
8). Why when analysing with Qdec I get more clusters? Are this defining or representing the sig.mgh as in GLM?
9). When using Qdec were I can find as an output of results the number of clusters and the cortical thickness value?
Thank you and have a great day! Antonella
Dear FS experts,
I finished running the GLM and I wonder what is the best way to further analyse my data in order to see if there is any relation between age (at seizure onset) and cortical thickness for my patients versus control group.
1). Why when I overlay the sig.mgh in tksurfer lh inflated, I get more clusters than the number of clusters obtained from mri_glmfit-sim?
2). What sig.mgh represents?
3) What's the real significance of a cluster (how a cluster is formed)?
4). Why my thickness value in clusters (eg cluster no. 1 which coresponds to the posteriorcingulate) for subject no.1 is different that the thickness value obtained for the same region in the lh.aparc.stats while running recon-all?
5). After running the GLM should I use visualizing and plotting method to further analyse my data and load FSGD file lh.gender_age.glmdir/y.fsgd?
6). Should my ROI's be defined or be the same with my clusters?
7). What is the difference between GLM and Qdec? What method is the best to analyse the relation between age and cortical thickness for my patients versus control group?
8). Why when analysing with Qdec I get more clusters? Are this defining or representing the sig.mgh as in GLM?
9). When using Qdec were I can find as an output of results the number of clusters and the cortical thickness value?
Thank you and have a great day! Antonella
Antonella Kis wrote:
Dear FS experts,
I finished running the GLM and I wonder what is the best way to further analyse my data in order to see if there is any relation between age (at seizure onset) and cortical thickness for my patients versus control group.
1). Why when I overlay the sig.mgh in tksurfer lh inflated, I get more clusters than the number of clusters obtained from mri_glmfit-sim?
When you look at it in tksurfer, you are looking at uncorrected data. glmfit-sim corrects for multiple comparisons. Some of the clusters don't survive.
2). What sig.mgh represents?
-log10(p)
- What's the real significance of a cluster (how a cluster is formed)?
It is based on the likelihood of getting a cluster of that size by chance given the search space (cortical surface area), smoothness (FWHM), and cluster-forming threshold (eg, p<.05).
4). Why my thickness value in clusters (eg cluster no. 1 which coresponds to the posteriorcingulate) for subject no.1 is different that the thickness value obtained for the same region in the lh.aparc.stats while running recon-all?
5). After running the GLM should I use visualizing and plotting method to further analyse my data and load FSGD file lh.gender_age.glmdir/y.fsgd?
6). Should my ROI's be defined or be the same with my clusters?
I don't know what you mean by that.
7). What is the difference between GLM and Qdec? What method is the best to analyse the relation between age and cortical thickness for my patients versus control group?
They are the same statistically. They are just different ways to provide information about your design and contrasts. QDEC is graphical (point and click). mri_glmfit you create FSGD files and run it from the command-line. QDEC actually creates FSGD files and contrast files and runs mri_glmfit.
8). Why when analysing with Qdec I get more clusters? Are this defining or representing the sig.mgh as in GLM?
They should give identical results when run in the same way. My guess is that you have created two different designs and so are getting different answers.
9). When using Qdec were I can find as an output of results the number of clusters and the cortical thickness value?
You'll have to run the correction for multiple comparisons (interface on the results page). This simply runs mri_glmfit-sim.
doug
Thank you and have a great day! Antonella
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Hi Doug,
I would like to run the correction for multiple comparisons. I know glmfit-sim corrects for multiple comparisons. My question is:
the corrections for multiple comparisons will be done while running
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --overwrite
or I need to add --cwpvalthresh .025 before --overwrite. I found on-line that setting a treshold = 2 means p < 0.05 but adding --cwpvalthresh .025 will do the correction across 2 spaces: lh and rh: .025 = .05/2 Bonferroni Correction and will show only the clusters with p<.025.
Is this right? So to run for multiple corrections in fact I will need to run:
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --cwpvalthresh .025 --overwrite
Please advise me.
Thank you for your time and help. Antonella
________________________________ From: Douglas N Greve greve@nmr.mgh.harvard.edu To: Antonella Kis atorok9@yahoo.com Cc: "freesurfer@nmr.mgh.harvard.edu" freesurfer@nmr.mgh.harvard.edu Sent: Monday, August 22, 2011 2:28 PM Subject: Re: [Freesurfer] GLM or Qdec ?
Antonella Kis wrote:
Dear FS experts,
I finished running the GLM and I wonder what is the best way to further analyse my data in order to see if there is any relation between age (at seizure onset) and cortical thickness for my patients versus control group.
1). Why when I overlay the sig.mgh in tksurfer lh inflated, I get more clusters than the number of clusters obtained from mri_glmfit-sim?
When you look at it in tksurfer, you are looking at uncorrected data. glmfit-sim corrects for multiple comparisons. Some of the clusters don't survive.
2). What sig.mgh represents?
-log10(p)
- What's the real significance of a cluster (how a cluster is formed)?
It is based on the likelihood of getting a cluster of that size by chance given the search space (cortical surface area), smoothness (FWHM), and cluster-forming threshold (eg, p<.05).
4). Why my thickness value in clusters (eg cluster no. 1 which coresponds to the posteriorcingulate) for subject no.1 is different that the thickness value obtained for the same region in the lh.aparc.stats while running recon-all?
5). After running the GLM should I use visualizing and plotting method to further analyse my data and load FSGD file lh.gender_age.glmdir/y.fsgd?
6). Should my ROI's be defined or be the same with my clusters?
I don't know what you mean by that.
7). What is the difference between GLM and Qdec? What method is the best to analyse the relation between age and cortical thickness for my patients versus control group?
They are the same statistically. They are just different ways to provide information about your design and contrasts. QDEC is graphical (point and click). mri_glmfit you create FSGD files and run it from the command-line. QDEC actually creates FSGD files and contrast files and runs mri_glmfit.
8). Why when analysing with Qdec I get more clusters? Are this defining or representing the sig.mgh as in GLM?
They should give identical results when run in the same way. My guess is that you have created two different designs and so are getting different answers.
9). When using Qdec were I can find as an output of results the number of clusters and the cortical thickness value?
You'll have to run the correction for multiple comparisons (interface on the results page). This simply runs mri_glmfit-sim.
doug
Thank you and have a great day! Antonella
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Yes, that is correct, you should use .025 if you are correcting for both hemispheres. doug
Antonella Kis wrote:
Hi Doug,
I would like to run the correction for multiple comparisons. I know glmfit-sim corrects for multiple comparisons. My question is:
the corrections for multiple comparisons will be done while running
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --overwrite
or I need to add --cwpvalthresh .025 before --overwrite. I found on-line that setting a treshold = 2 means p < 0.05 but adding --cwpvalthresh .025 will do the correction across 2 spaces: lh and rh: .025 = .05/2 Bonferroni Correction and will show only the clusters with p<.025.
Is this right? So to run for multiple corrections in fact I will need to run:
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --cwpvalthresh .025 --overwrite
Please advise me.
Thank you for your time and help. Antonella
*From:* Douglas N Greve greve@nmr.mgh.harvard.edu *To:* Antonella Kis atorok9@yahoo.com *Cc:* "freesurfer@nmr.mgh.harvard.edu" freesurfer@nmr.mgh.harvard.edu *Sent:* Monday, August 22, 2011 2:28 PM *Subject:* Re: [Freesurfer] GLM or Qdec ?
Antonella Kis wrote:
Dear FS experts,
I finished running the GLM and I wonder what is the best way to further analyse my data in order to see if there is any relation between age (at seizure onset) and cortical thickness for my patients versus control group.
1). Why when I overlay the sig.mgh in tksurfer lh inflated, I get more clusters than the number of clusters obtained from mri_glmfit-sim?
When you look at it in tksurfer, you are looking at uncorrected data. glmfit-sim corrects for multiple comparisons. Some of the clusters don't survive.
2). What sig.mgh represents?
-log10(p)
- What's the real significance of a cluster (how a cluster is formed)?
It is based on the likelihood of getting a cluster of that size by chance given the search space (cortical surface area), smoothness (FWHM), and cluster-forming threshold (eg, p<.05).
4). Why my thickness value in clusters (eg cluster no. 1 which coresponds to the posteriorcingulate) for subject no.1 is different that the thickness value obtained for the same region in the lh.aparc.stats while running recon-all?
5). After running the GLM should I use visualizing and plotting method to further analyse my data and load FSGD file lh.gender_age.glmdir/y.fsgd?
6). Should my ROI's be defined or be the same with my clusters?
I don't know what you mean by that.
7). What is the difference between GLM and Qdec? What method is the best to analyse the relation between age and cortical thickness for my patients versus control group?
They are the same statistically. They are just different ways to provide information about your design and contrasts. QDEC is graphical (point and click). mri_glmfit you create FSGD files and run it from the command-line. QDEC actually creates FSGD files and contrast files and runs mri_glmfit.
8). Why when analysing with Qdec I get more clusters? Are this defining or representing the sig.mgh as in GLM?
They should give identical results when run in the same way. My guess is that you have created two different designs and so are getting different answers.
9). When using Qdec were I can find as an output of results the number of clusters and the cortical thickness value?
You'll have to run the correction for multiple comparisons (interface on the results page). This simply runs mri_glmfit-sim.
doug
Thank you and have a great day! Antonella
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu mailto:Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu mailto:greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422
Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu mailto:Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.
Good morning Doug,
Thank you for your help. I followed your advise and unfortunately I do not understand why after adding the threshold 2 cwpvalthresh .025 in order to do the correction across lh and rh I still get in my clusters summary and clusters with cwp>.01 as well as cwp> .025. This is what I was running:
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --cwpvalthresh .025 \ --overwrite
And this is my output:
# ClusterNo Max VtxMax Size(mm^2) MNIX MNIY MNIZ CWP CWPLow CWPHi NVtxs Annot
1 -6.24 104279 2004.41 6.4 -83.6 34.4 0.0001 0 0 3539 cuneus
2 -5.71 56656 2923.03 24.4 25.4 37.1 0.0001 0 0 5020 superiorfrontal
3 -4.89 156133 1826.82 5.1 -29.7 38.6 0.0001 0 0 4324 posteriorcingulate
4 -4.57 117278 1917.42 43.2 43.4 -8.2 0.0001 0 0 3393 parsorbitalis
5 -4.32 134603 910.13 10.5 45.7 -9.2 0.0009 0 0 1597 medialorbitofrontal
6 -4.01 47845 1018.06 38.9 -73.9 34.6 0.0003 0 0 1802 inferiorparietal
7 -3.97 12268 580.81 14.2 -51 63.3 0.0132 0.01 0.01 1415 superiorparietal
8 -3.78 54798 871.02 53 -39.7 -5.1 0.0010 0 0 1969 bankssts
9 -3.78 21941 555.51 45.8 -28.2 6.4 0.0175 0.02 0.02 1369 superiortemporal
10 -2.91 86118 884.25 10.1 -98.9 11.7 0.0010 0 0 1089 lateraloccipital
11 -2.86 13271 674.26 21.5 -65.4 -6.1 0.0056 0 0.01 722 lingual
Is this right?
Please advise.
Thank you. Antonella
________________________________ From: Douglas N Greve greve@nmr.mgh.harvard.edu To: Antonella Kis atorok9@yahoo.com Cc: "freesurfer@nmr.mgh.harvard.edu" freesurfer@nmr.mgh.harvard.edu Sent: Tuesday, August 23, 2011 4:33 PM Subject: Re: [Freesurfer] correction for multiple comparisons
Yes, that is correct, you should use .025 if you are correcting for both hemispheres. doug
Antonella Kis wrote:
Hi Doug,
I would like to run the correction for multiple comparisons. I know glmfit-sim corrects for multiple comparisons. My question is:
the corrections for multiple comparisons will be done while running
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --overwrite
or I need to add --cwpvalthresh .025 before --overwrite. I found on-line that setting a treshold = 2 means p < 0.05 but adding --cwpvalthresh .025 will do the correction across 2 spaces: lh and rh: .025 = .05/2 Bonferroni Correction and will show only the clusters with p<.025. Is this right? So to run for multiple corrections in fact I will need to run:
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --cwpvalthresh .025 --overwrite
Please advise me.
Thank you for your time and help. Antonella
*From:* Douglas N Greve greve@nmr.mgh.harvard.edu *To:* Antonella Kis atorok9@yahoo.com *Cc:* "freesurfer@nmr.mgh.harvard.edu" freesurfer@nmr.mgh.harvard.edu *Sent:* Monday, August 22, 2011 2:28 PM *Subject:* Re: [Freesurfer] GLM or Qdec ?
Antonella Kis wrote:
Dear FS experts,
I finished running the GLM and I wonder what is the best way to further analyse my data in order to see if there is any relation between age (at seizure onset) and cortical thickness for my patients versus control group.
1). Why when I overlay the sig.mgh in tksurfer lh inflated, I get more clusters than the number of clusters obtained from mri_glmfit-sim?
When you look at it in tksurfer, you are looking at uncorrected data. glmfit-sim corrects for multiple comparisons. Some of the clusters don't survive.
2). What sig.mgh represents?
-log10(p)
- What's the real significance of a cluster (how a cluster is formed)?
It is based on the likelihood of getting a cluster of that size by chance given the search space (cortical surface area), smoothness (FWHM), and cluster-forming threshold (eg, p<.05).
4). Why my thickness value in clusters (eg cluster no. 1 which coresponds to the posteriorcingulate) for subject no.1 is different that the thickness value obtained for the same region in the lh.aparc.stats while running recon-all?
5). After running the GLM should I use visualizing and plotting method to further analyse my data and load FSGD file lh.gender_age.glmdir/y.fsgd?
6). Should my ROI's be defined or be the same with my clusters?
I don't know what you mean by that.
7). What is the difference between GLM and Qdec? What method is the best to analyse the relation between age and cortical thickness for my patients versus control group?
They are the same statistically. They are just different ways to provide information about your design and contrasts. QDEC is graphical (point and click). mri_glmfit you create FSGD files and run it from the command-line. QDEC actually creates FSGD files and contrast files and runs mri_glmfit.
8). Why when analysing with Qdec I get more clusters? Are this defining or representing the sig.mgh as in GLM?
They should give identical results when run in the same way. My guess is that you have created two different designs and so are getting different answers.
9). When using Qdec were I can find as an output of results the number of clusters and the cortical thickness value?
You'll have to run the correction for multiple comparisons (interface on the results page). This simply runs mri_glmfit-sim.
doug
Thank you and have a great day! Antonella
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu mailto:Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu mailto:greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422
Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu mailto:Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.
Hi Bruce,
Thank you very much for your prompt and helpful answer.
I have one more question, please: I do not understand what's the role of adding the threshold value while running the mri_glmfit-sim with --cache since after adding the threshold 2, I still get clusters with cwp>.01.
This is what I was running:
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --cwpvalthresh .025 \ --overwrite
And this is my output:
# ClusterNo Max VtxMax Size(mm^2) MNIX MNIY MNIZ CWP CWPLow CWPHi NVtxs Annot
1 -6.24 104279 2004.41 6.4 -83.6 34.4 0.0001 0 0 3539 cuneus
2 -5.71 56656 2923.03 24.4 25.4 37.1 0.0001 0 0 5020 superiorfrontal
3 -4.89 156133 1826.82 5.1 -29.7 38.6 0.0001 0 0 4324 posteriorcingulate
4 -4.57 117278 1917.42 43.2 43.4 -8.2 0.0001 0 0 3393 parsorbitalis
5 -4.32 134603 910.13 10.5 45.7 -9.2 0.0009 0 0 1597 medialorbitofrontal
6 -4.01 47845 1018.06 38.9 -73.9 34.6 0.0003 0 0 1802 inferiorparietal
7 -3.97 12268 580.81 14.2 -51 63.3 0.0132 0.01 0.01 1415 superiorparietal
8 -3.78 54798 871.02 53 -39.7 -5.1 0.0010 0 0 1969 bankssts
9 -3.78 21941 555.51 45.8 -28.2 6.4 0.0175 0.02 0.02 1369 superiortemporal
10 -2.91 86118 884.25 10.1 -98.9 11.7 0.0010 0 0 1089 lateraloccipital
11 -2.86 13271 674.26 21.5 -65.4 -6.1 0.0056 0 0.01 722 lingual
Is this right? Also, by adding for the lh --cwpvalthresh .025 what exactly means or how the corrections are made for both hemispheres. I know --cwpvalthresh .025 shows only clusters p<.025 but then what means cluster significance values not corrected for 2 spaces?
Please advise.
Thank you. Antonella
________________________________ From: Douglas N Greve greve@nmr.mgh.harvard.edu To: Antonella Kis atorok9@yahoo.com Cc: "freesurfer@nmr.mgh.harvard.edu" freesurfer@nmr.mgh.harvard.edu Sent: Tuesday, August 23, 2011 4:33 PM Subject: Re: [Freesurfer] correction for multiple comparisons
Yes, that is correct, you should use .025 if you are correcting for both hemispheres. doug
Antonella Kis wrote:
Hi Doug,
I would like to run the correction for multiple
comparisons. I know
glmfit-sim corrects for multiple comparisons. My question is:
the corrections for multiple comparisons will be done while running
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --overwrite
or I need to add --cwpvalthresh .025 before --overwrite. I found on-line that setting a treshold = 2 means p < 0.05 but adding --cwpvalthresh .025 will do the correction across 2 spaces: lh and rh: .025 = .05/2 Bonferroni Correction and will show only the clusters with p<.025. Is this right? So to run for multiple corrections in fact I will need to run:
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --cwpvalthresh .025 --overwrite
Please advise me.
Thank you for your time and
help.
Antonella
*From:* Douglas N Greve greve@nmr.mgh.harvard.edu *To:* Antonella Kis atorok9@yahoo.com *Cc:* "freesurfer@nmr.mgh.harvard.edu" freesurfer@nmr.mgh.harvard.edu *Sent:* Monday, August 22, 2011 2:28 PM *Subject:* Re: [Freesurfer] GLM or Qdec ?
Antonella Kis wrote:
Dear FS experts,
I finished running the GLM and I wonder what is the best way to further
analyse my data in order to see if there is any relation
between age (at seizure onset) and cortical thickness for my patients versus control group.
1). Why when I overlay the sig.mgh in tksurfer lh inflated, I get more clusters than the number of clusters obtained from mri_glmfit-sim?
When you look at it in tksurfer, you are looking at uncorrected data. glmfit-sim corrects for multiple comparisons. Some of the clusters don't survive.
2). What sig.mgh represents?
-log10(p)
- What's the real significance of a cluster (how a cluster is formed)?
It is based on the likelihood of getting a cluster of that size by chance given the search space (cortical surface area), smoothness (FWHM), and cluster-forming threshold (eg, p<.05).
4). Why my thickness value in clusters (eg cluster no. 1 which
coresponds to the posteriorcingulate) for subject no.1 is different that the thickness value obtained for the same region in the lh.aparc.stats while running recon-all?
5). After running the GLM should I use visualizing and plotting method to further analyse my data and load FSGD file lh.gender_age.glmdir/y.fsgd?
6). Should my ROI's be defined or be the same with my clusters?
I don't know what you mean by that.
7). What is the difference between GLM and Qdec? What method is the best to analyse the relation between age and cortical thickness for my patients versus control group?
They are the same statistically. They are just different ways to provide information about your design
and contrasts. QDEC is graphical (point
and click). mri_glmfit you create FSGD files and run it from the command-line. QDEC actually creates FSGD files and contrast files and runs mri_glmfit.
8). Why when analysing with Qdec I get more clusters? Are this defining or representing the sig.mgh as in GLM?
They should give identical results when run in the same way. My guess is that you have created two different designs and so are getting different answers.
9). When using Qdec were I can find as an output of results the number of clusters and the cortical thickness value?
You'll have to run the correction for multiple comparisons (interface on the results page). This simply runs mri_glmfit-sim.
doug
Thank you and have a great day!
Antonella
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu mailto:Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu mailto:greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422
Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu mailto:Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
The information in this e-mail is intended only for the person to whom it is addressed. If you believe this
e-mail was sent to you in error and the
e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.
I'll leave that for Doug On Thu, 25 Aug 2011, Antonella Kis wrote:
Hi Bruce,
Thank you very much for your prompt and helpful answer.
I have one more question, please: I do not understand what's the role of adding the threshold value while running the mri_glmfit-sim with --cache since after adding the threshold 2, I still get clusters with cwp>.01. This is what I was running:
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --cwpvalthresh .025 \ --overwrite
And this is my output:
# ClusterNo Max VtxMax Size(mm^2) MNIX MNIY MNIZ CWP CWPLow CWPHi NVtxs Annot
1 -6.24 104279 2004.41 6.4 -83.6 34.4 0.0001 0 0 3539 cuneus
2 -5.71 56656 2923.03 24.4 25.4 37.1 0.0001 0 0 5020 superiorfrontal
3 -4.89 156133 1826.82 5.1 -29.7 38.6 0.0001 0 0 4324 posteriorcingulate
4 -4.57 117278 1917.42 43.2 43.4 -8.2 0.0001 0 0 3393 parsorbitalis
5 -4.32 134603 910.13 10.5 45.7 -9.2 0.0009 0 0 1597 medialorbitofrontal
6 -4.01 47845 1018.06 38.9 -73.9 34.6 0.0003 0 0 1802 inferiorparietal
7 -3.97 12268 580.81 14.2 -51 63.3 0.0132 0.01 0.01 1415 superiorparietal
8 -3.78 54798 871.02 53 -39.7 -5.1 0.0010 0 0 1969 bankssts
9 -3.78 21941 555.51 45.8 -28.2 6.4 0.0175 0.02 0.02 1369 superiortemporal
10 -2.91 86118 884.25 10.1 -98.9 11.7 0.0010 0 0 1089 lateraloccipital
11 -2.86 13271 674.26 21.5 -65.4 -6.1 0.0056 0 0.01 722 lingual
Is this right?
Also, by adding for the lh --cwpvalthresh .025 what exactly means or how the corrections are made for both hemispheres. I know --cwpvalthresh .025 shows only clusters p<.025 but then what means cluster significance values not corrected for 2 spaces? Please advise.
Thank you. Antonella
From: Douglas N Greve greve@nmr.mgh.harvard.edu To: Antonella Kis atorok9@yahoo.com Cc: "freesurfer@nmr.mgh.harvard.edu" freesurfer@nmr.mgh.harvard.edu Sent: Tuesday, August 23, 2011 4:33 PM Subject: Re: [Freesurfer] correction for multiple comparisons
Yes, that is correct, you should use .025 if you are correcting for both hemispheres. doug
Antonella Kis wrote:
Hi Doug,
I would like to run the correction for multiple comparisons. I know glmfit-sim corrects for multiple comparisons. My question is:
the corrections for multiple comparisons will be done while running
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --overwrite
or I need to add --cwpvalthresh .025 before --overwrite. I found on-line that setting a treshold = 2 means p < 0.05 but adding --cwpvalthresh .025 will do the correction across 2 spaces: lh and rh: .025 = .05/2 Bonferroni Correction and will show only the clusters with p<.025. Is this right? So to run for multiple corrections in fact I will need to run:
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --cwpvalthresh .025 --overwrite
Please advise me.
Thank you for your time and help. Antonella
*From:* Douglas N Greve greve@nmr.mgh.harvard.edu *To:* Antonella Kis atorok9@yahoo.com *Cc:* "freesurfer@nmr.mgh.harvard.edu" freesurfer@nmr.mgh.harvard.edu *Sent:* Monday, August 22, 2011 2:28 PM *Subject:* Re: [Freesurfer] GLM or Qdec ?
Antonella Kis wrote:
Dear FS experts,
I finished running the GLM and I wonder what is the best way to further analyse my data in order to see if there is any relation between age (at seizure onset) and cortical thickness for my patients versus control group.
1). Why when I overlay the sig.mgh in tksurfer lh inflated, I get more clusters than the number of clusters obtained from mri_glmfit-sim?
When you look at it in tksurfer, you are looking at uncorrected data. glmfit-sim corrects for multiple comparisons. Some of the clusters don't survive.
2). What sig.mgh represents?
-log10(p)
- What's the real significance of a cluster (how a cluster is formed)?
It is based on the likelihood of getting a cluster of that size by chance given the search space (cortical surface area), smoothness (FWHM), and cluster-forming threshold (eg, p<.05).
4). Why my thickness value in clusters (eg cluster no. 1 which coresponds to the posteriorcingulate) for subject no.1 is different that the thickness value obtained for the same region in the lh.aparc.stats while running recon-all?
5). After running the GLM should I use visualizing and plotting method to further analyse my data and load FSGD file lh.gender_age.glmdir/y.fsgd?
6). Should my ROI's be defined or be the same with my clusters?
I don't know what you mean by that.
7). What is the difference between GLM and Qdec? What method is the best to analyse the relation between age and cortical thickness for my patients versus control group?
They are the same statistically. They are just different ways to provide information about your design and contrasts. QDEC is graphical (point and click). mri_glmfit you create FSGD files and run it from the command-line. QDEC actually creates FSGD files and contrast files and runs mri_glmfit.
8). Why when analysing with Qdec I get more clusters? Are this defining or representing the sig.mgh as in GLM?
They should give identical results when run in the same way. My guess is that you have created two different designs and so are getting different answers.
9). When using Qdec were I can find as an output of results the number of clusters and the cortical thickness value?
You'll have to run the correction for multiple comparisons (interface on the results page). This simply runs mri_glmfit-sim.
doug
Thank you and have a great day! Antonella
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu mailto:Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu mailto:greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422
Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu mailto:Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.
-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422
Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
That "2" is a voxel-wise threshold used to form the cluster. The cwp is a cluster-wise threshold used to eliminate non-significant clusters from the output.
Bruce Fischl wrote:
I'll leave that for Doug On Thu, 25 Aug 2011, Antonella Kis wrote:
Hi Bruce,
Thank you very much for your prompt and helpful answer.
I have one more question, please: I do not understand what's the role of adding the threshold value while running the mri_glmfit-sim with --cache since after adding the threshold 2, I still get clusters with cwp>.01. This is what I was running:
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --cwpvalthresh .025 \ --overwrite
And this is my output:
# ClusterNo Max VtxMax Size(mm^2) MNIX MNIY MNIZ CWP CWPLow CWPHi NVtxs Annot
1 -6.24 104279 2004.41 6.4 -83.6 34.4 0.0001 0 0 3539 cuneus
2 -5.71 56656 2923.03 24.4 25.4 37.1 0.0001 0 0 5020 superiorfrontal
3 -4.89 156133 1826.82 5.1 -29.7 38.6 0.0001 0 0 4324 posteriorcingulate
4 -4.57 117278 1917.42 43.2 43.4 -8.2 0.0001 0 0 3393 parsorbitalis
5 -4.32 134603 910.13 10.5 45.7 -9.2 0.0009 0 0 1597 medialorbitofrontal
6 -4.01 47845 1018.06 38.9 -73.9 34.6 0.0003 0 0 1802 inferiorparietal
7 -3.97 12268 580.81 14.2 -51 63.3 0.0132 0.01 0.01 1415 superiorparietal
8 -3.78 54798 871.02 53 -39.7 -5.1 0.0010 0 0 1969 bankssts
9 -3.78 21941 555.51 45.8 -28.2 6.4 0.0175 0.02 0.02 1369 superiortemporal
10 -2.91 86118 884.25 10.1 -98.9 11.7 0.0010 0 0 1089 lateraloccipital
11 -2.86 13271 674.26 21.5 -65.4 -6.1 0.0056 0 0.01 722 lingual
Is this right?
Also, by adding for the lh --cwpvalthresh .025 what exactly means or how the corrections are made for both hemispheres. I know --cwpvalthresh .025 shows only clusters p<.025 but then what means cluster significance values not corrected for 2 spaces? Please advise.
Thank you. Antonella
From: Douglas N Greve greve@nmr.mgh.harvard.edu To: Antonella Kis atorok9@yahoo.com Cc: "freesurfer@nmr.mgh.harvard.edu" freesurfer@nmr.mgh.harvard.edu Sent: Tuesday, August 23, 2011 4:33 PM Subject: Re: [Freesurfer] correction for multiple comparisons
Yes, that is correct, you should use .025 if you are correcting for both hemispheres. doug
Antonella Kis wrote:
Hi Doug,
I would like to run the correction for multiple comparisons. I know glmfit-sim corrects for multiple comparisons. My question is:
the corrections for multiple comparisons will be done while running
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --overwrite
or I need to add --cwpvalthresh .025 before --overwrite. I found on-line that setting a treshold = 2 means p < 0.05 but adding --cwpvalthresh .025 will do the correction across 2 spaces: lh and rh: .025 = .05/2 Bonferroni Correction and will show only the clusters with p<.025.
Is this right? So to run for multiple corrections in fact I will need to run:
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --cwpvalthresh .025 --overwrite
Please advise me.
Thank you for your time and help. Antonella
*From:* Douglas N Greve greve@nmr.mgh.harvard.edu *To:* Antonella Kis atorok9@yahoo.com *Cc:* "freesurfer@nmr.mgh.harvard.edu"
freesurfer@nmr.mgh.harvard.edu
*Sent:* Monday, August 22, 2011 2:28 PM *Subject:* Re: [Freesurfer] GLM or Qdec ?
Antonella Kis wrote:
Dear FS experts,
I finished running the GLM and I wonder what is the best way to further analyse my data in order to see if there is any relation between age (at seizure onset) and cortical thickness for my
patients
versus control group.
1). Why when I overlay the sig.mgh in tksurfer lh inflated, I get more clusters than the number of clusters obtained from
mri_glmfit-sim?
When you look at it in tksurfer, you are looking at uncorrected data. glmfit-sim corrects for multiple comparisons. Some of the clusters
don't
survive.
2). What sig.mgh represents?
-log10(p)
- What's the real significance of a cluster (how a cluster is
formed)?
It is based on the likelihood of getting a cluster of that size by chance given the search space (cortical surface area), smoothness (FWHM), and cluster-forming threshold (eg, p<.05).
4). Why my thickness value in clusters (eg cluster no. 1 which coresponds to the posteriorcingulate) for subject no.1 is different that the thickness value obtained for the same region in the lh.aparc.stats while running recon-all?
5). After running the GLM should I use visualizing and plotting
method
to further analyse my data and load FSGD file lh.gender_age.glmdir/y.fsgd?
6). Should my ROI's be defined or be the same with my clusters?
I don't know what you mean by that.
7). What is the difference between GLM and Qdec? What method is the best to analyse the relation between age and cortical thickness for my patients versus control group?
They are the same statistically. They are just different ways to
provide
information about your design and contrasts. QDEC is graphical (point and click). mri_glmfit you create FSGD files and run it from the command-line. QDEC actually creates FSGD files and contrast files and runs mri_glmfit.
8). Why when analysing with Qdec I get more clusters? Are this defining or representing the sig.mgh as in GLM?
They should give identical results when run in the same way. My
guess is
that you have created two different designs and so are getting
different
answers.
9). When using Qdec were I can find as an output of results the number of clusters and the cortical thickness value?
You'll have to run the correction for multiple comparisons
(interface on
the results page). This simply runs mri_glmfit-sim.
doug
Thank you and have a great day! Antonella
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu
mailto:Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu mailto:greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422
Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu mailto:Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.
-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422
Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Hi Doug,
Many thanks for your valuable answer. Can you please briefly explain some other issues:
1). what exactly means Correction for Multiple Comparisons and how this is done when using --sim method with --cache.
2). do I need to do an estimation FDRthresholding in tksurfer for GLM method before running the --sim?
3).I know that adding cwpvalthresh .025 shows only clusters p<.025 but then what means cluster significance values not corrected for 2 spaces? I found this on the slides posted on-line.
Many thanks and have a great day! Antonella
________________________________
Hi Doug,
Since running mri_glmfit-sim will do a correction by multiple comparison (I suppose is doing a Monte Carlo to remove areas that are not statically significant) I wonder why I still get clusters with p>0.01? Do I need to apply FDR before -sim?
Thank you for your time and help. Antonella
________________________________ From: Douglas N Greve greve@nmr.mgh.harvard.edu To: Bruce Fischl fischl@nmr.mgh.harvard.edu Cc: Antonella Kis atorok9@yahoo.com; "freesurfer@nmr.mgh.harvard.edu" freesurfer@nmr.mgh.harvard.edu Sent: Thursday, August 25, 2011 12:23 PM Subject: Re: [Freesurfer] Fw: fsaverage
That "2" is a voxel-wise threshold used to form the cluster. The cwp is a cluster-wise threshold used to eliminate non-significant clusters from the output.
Bruce Fischl wrote:
I'll leave that for Doug On Thu, 25 Aug 2011, Antonella Kis wrote:
Hi Bruce,
Thank you very much for your prompt and helpful answer.
I have one more question, please: I do not understand what's the role of adding the threshold value while running the mri_glmfit-sim with --cache since after adding the threshold 2, I still get clusters with cwp>.01. This is what I was running:
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --cwpvalthresh .025 \ --overwrite
And this is my output:
# ClusterNo Max VtxMax Size(mm^2) MNIX MNIY MNIZ CWP CWPLow CWPHi NVtxs Annot
1 -6.24 104279 2004.41 6.4 -83.6 34.4 0.0001 0 0 3539 cuneus
2 -5.71 56656 2923.03 24.4 25.4 37.1 0.0001 0 0 5020 superiorfrontal
3 -4.89 156133 1826.82 5.1 -29.7 38.6 0.0001 0 0 4324 posteriorcingulate
4 -4.57 117278 1917.42 43.2 43.4 -8.2 0.0001 0 0 3393 parsorbitalis
5 -4.32 134603 910.13 10.5 45.7 -9.2 0.0009 0 0 1597 medialorbitofrontal
6 -4.01 47845 1018.06 38.9 -73.9 34.6 0.0003 0 0 1802 inferiorparietal
7 -3.97 12268 580.81 14.2 -51 63.3 0.0132 0.01 0.01 1415 superiorparietal
8 -3.78 54798 871.02 53 -39.7 -5.1 0.0010 0 0 1969 bankssts
9 -3.78 21941 555.51 45.8 -28.2 6.4 0.0175 0.02 0.02 1369 superiortemporal
10 -2.91 86118 884.25 10.1 -98.9 11.7 0.0010 0 0 1089 lateraloccipital
11 -2.86 13271 674.26 21.5 -65.4 -6.1 0.0056 0 0.01 722 lingual
Is this right?
Also, by adding for the lh --cwpvalthresh .025 what exactly means or how the corrections are made for both hemispheres. I know --cwpvalthresh .025 shows only clusters p<.025 but then what means cluster significance values not corrected for 2 spaces? Please advise.
Thank you. Antonella
From: Douglas N Greve greve@nmr.mgh.harvard.edu To: Antonella Kis atorok9@yahoo.com Cc: "freesurfer@nmr.mgh.harvard.edu" freesurfer@nmr.mgh.harvard.edu Sent: Tuesday, August 23, 2011 4:33 PM Subject: Re: [Freesurfer] correction for multiple comparisons
Yes, that is correct, you should use .025 if you are correcting for both hemispheres. doug
Antonella Kis wrote:
Hi Doug,
I would like to run the correction for multiple comparisons. I know glmfit-sim corrects for multiple comparisons. My question is:
the corrections for multiple comparisons will be done while running
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --overwrite
or I need to add --cwpvalthresh .025 before --overwrite. I found on-line that setting a treshold = 2 means p < 0.05 but adding --cwpvalthresh .025 will do the correction across 2 spaces: lh and rh: .025 = .05/2 Bonferroni Correction and will show only the clusters with p<.025.
Is this right? So to run for multiple corrections in fact I will need to run:
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --cwpvalthresh .025 --overwrite
Please advise me.
Thank you for your time and help. Antonella
*From:* Douglas N Greve greve@nmr.mgh.harvard.edu *To:* Antonella Kis atorok9@yahoo.com *Cc:* "freesurfer@nmr.mgh.harvard.edu"
freesurfer@nmr.mgh.harvard.edu
*Sent:* Monday, August 22, 2011 2:28 PM *Subject:* Re: [Freesurfer] GLM or Qdec ?
Antonella Kis wrote:
Dear FS experts,
I finished running the GLM and I wonder what is the best way to further analyse my data in order to see if there is any relation between age (at seizure onset) and cortical thickness for my
patients
versus control group.
1). Why when I overlay the sig.mgh in tksurfer lh inflated, I get more clusters than the number of clusters obtained from
mri_glmfit-sim?
When you look at it in tksurfer, you are looking at uncorrected data. glmfit-sim corrects for multiple comparisons. Some of the clusters
don't
survive.
2). What sig.mgh represents?
-log10(p)
- What's the real significance of a cluster (how a cluster is
formed)?
It is based on the likelihood of getting a cluster of that size by chance given the search space (cortical surface area), smoothness (FWHM), and cluster-forming threshold (eg, p<.05).
4). Why my thickness value in clusters (eg cluster no. 1 which coresponds to the posteriorcingulate) for subject no.1 is different that the thickness value obtained for the same region in the lh.aparc.stats while running recon-all?
5). After running the GLM should I use visualizing and plotting
method
to further analyse my data and load FSGD file lh.gender_age.glmdir/y.fsgd?
6). Should my ROI's be defined or be the same with my clusters?
I don't know what you mean by that.
7). What is the difference between GLM and Qdec? What method is the best to analyse the relation between age and cortical thickness for my patients versus control group?
They are the same statistically. They are just different ways to
provide
information about your design and contrasts. QDEC is graphical (point and click). mri_glmfit you create FSGD files and run it from the command-line. QDEC actually creates FSGD files and contrast files and runs mri_glmfit.
8). Why when analysing with Qdec I get more clusters? Are this defining or representing the sig.mgh as in GLM?
They should give identical results when run in the same way. My
guess is
that you have created two different designs and so are getting
different
answers.
9). When using Qdec were I can find as an output of results the number of clusters and the cortical thickness value?
You'll have to run the correction for multiple comparisons
(interface on
the results page). This simply runs mri_glmfit-sim.
doug
Thank you and have a great day! Antonella
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu
mailto:Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu mailto:greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422
Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu mailto:Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.
-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422
Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Are you setting the cwp to .01? The cwp controls the threshold for cluster significance. The number after "--cache" controls the cluster-forming threshold
Antonella Kis wrote:
Hi Doug,
Since running mri_glmfit-sim will do a correction by multiple comparison (I suppose is doing a Monte Carlo to remove areas that are not statically significant) I wonder why I still get clusters with p>0.01? Do I need to apply FDR before -sim?
Thank you for your time and help. Antonella
*From:* Douglas N Greve greve@nmr.mgh.harvard.edu *To:* Bruce Fischl fischl@nmr.mgh.harvard.edu *Cc:* Antonella Kis atorok9@yahoo.com; "freesurfer@nmr.mgh.harvard.edu" freesurfer@nmr.mgh.harvard.edu *Sent:* Thursday, August 25, 2011 12:23 PM *Subject:* Re: [Freesurfer] Fw: fsaverage
That "2" is a voxel-wise threshold used to form the cluster. The cwp is a cluster-wise threshold used to eliminate non-significant clusters from the output.
Bruce Fischl wrote:
I'll leave that for Doug On Thu, 25 Aug 2011, Antonella Kis wrote:
Hi Bruce,
Thank you very much for your prompt and helpful answer.
I have one more question, please: I do not understand what's the role of adding the threshold value while running the mri_glmfit-sim with --cache since after adding the threshold 2, I still get clusters with cwp>.01. This is what I was running:
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --cwpvalthresh .025 \ --overwrite
And this is my output:
# ClusterNo Max VtxMax Size(mm^2) MNIX MNIY MNIZ CWP CWPLow CWPHi NVtxs Annot
1 -6.24 104279 2004.41 6.4 -83.6 34.4 0.0001 0 0 3539 cuneus
2 -5.71 56656 2923.03 24.4 25.4 37.1 0.0001 0 0 5020 superiorfrontal
3 -4.89 156133 1826.82 5.1 -29.7 38.6 0.0001 0 0 4324 posteriorcingulate
4 -4.57 117278 1917.42 43.2 43.4 -8.2 0.0001 0 0 3393 parsorbitalis
5 -4.32 134603 910.13 10.5 45.7 -9.2 0.0009 0 0 1597 medialorbitofrontal
6 -4.01 47845 1018.06 38.9 -73.9 34.6 0.0003 0 0 1802 inferiorparietal
7 -3.97 12268 580.81 14.2 -51 63.3 0.0132 0.01 0.01 1415 superiorparietal
8 -3.78 54798 871.02 53 -39.7 -5.1 0.0010 0 0 1969 bankssts
9 -3.78 21941 555.51 45.8 -28.2 6.4 0.0175 0.02 0.02 1369 superiortemporal
10 -2.91 86118 884.25 10.1 -98.9 11.7 0.0010 0 0 1089 lateraloccipital
11 -2.86 13271 674.26 21.5 -65.4 -6.1 0.0056 0 0.01 722 lingual
Is this right?
Also, by adding for the lh --cwpvalthresh .025 what exactly means or how the corrections are made for both hemispheres. I know --cwpvalthresh .025 shows only clusters p<.025 but then what means cluster significance values not corrected for 2 spaces? Please advise.
Thank you. Antonella
From: Douglas N Greve <greve@nmr.mgh.harvard.edu
mailto:greve@nmr.mgh.harvard.edu>
To: Antonella Kis <atorok9@yahoo.com mailto:atorok9@yahoo.com> Cc: "freesurfer@nmr.mgh.harvard.edu
mailto:freesurfer@nmr.mgh.harvard.edu" <freesurfer@nmr.mgh.harvard.edu mailto:freesurfer@nmr.mgh.harvard.edu>
Sent: Tuesday, August 23, 2011 4:33 PM Subject: Re: [Freesurfer] correction for multiple comparisons
Yes, that is correct, you should use .025 if you are correcting for
both
hemispheres. doug
Antonella Kis wrote:
Hi Doug,
I would like to run the correction for multiple comparisons. I know glmfit-sim corrects for multiple comparisons. My question is:
the corrections for multiple comparisons will be done while running
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --overwrite
or I need to add --cwpvalthresh .025 before --overwrite. I found on-line that setting a treshold = 2 means p < 0.05 but adding --cwpvalthresh .025 will do the correction across 2 spaces: lh
and rh:
.025 = .05/2 Bonferroni Correction and will show only the clusters with p<.025.
Is this right? So to run for multiple corrections in fact I will need to run:
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --cwpvalthresh .025 --overwrite
Please advise me.
Thank you for your time and help. Antonella
*From:* Douglas N Greve <greve@nmr.mgh.harvard.edu
mailto:greve@nmr.mgh.harvard.edu>
*To:* Antonella Kis <atorok9@yahoo.com mailto:atorok9@yahoo.com> *Cc:* "freesurfer@nmr.mgh.harvard.edu
mailto:freesurfer@nmr.mgh.harvard.edu"
<freesurfer@nmr.mgh.harvard.edu
mailto:freesurfer@nmr.mgh.harvard.edu>
*Sent:* Monday, August 22, 2011 2:28 PM *Subject:* Re: [Freesurfer] GLM or Qdec ?
Antonella Kis wrote:
Dear FS experts,
I finished running the GLM and I wonder what is the best way to further analyse my data in order to see if there is any relation between age (at seizure onset) and cortical thickness for my
patients
versus control group.
1). Why when I overlay the sig.mgh in tksurfer lh inflated, I get more clusters than the number of clusters obtained from
mri_glmfit-sim?
When you look at it in tksurfer, you are looking at uncorrected data. glmfit-sim corrects for multiple comparisons. Some of the clusters
don't
survive.
2). What sig.mgh represents?
-log10(p)
- What's the real significance of a cluster (how a cluster is
formed)?
It is based on the likelihood of getting a cluster of that size by chance given the search space (cortical surface area), smoothness (FWHM), and cluster-forming threshold (eg, p<.05).
4). Why my thickness value in clusters (eg cluster no. 1 which coresponds to the posteriorcingulate) for subject no.1 is different that the thickness value obtained for the same region in the lh.aparc.stats while running recon-all?
5). After running the GLM should I use visualizing and plotting
method
to further analyse my data and load FSGD file lh.gender_age.glmdir/y.fsgd?
6). Should my ROI's be defined or be the same with my clusters?
I don't know what you mean by that.
7). What is the difference between GLM and Qdec? What method is the best to analyse the relation between age and cortical
thickness for
my patients versus control group?
They are the same statistically. They are just different ways to
provide
information about your design and contrasts. QDEC is graphical (point and click). mri_glmfit you create FSGD files and run it from the command-line. QDEC actually creates FSGD files and contrast files and runs mri_glmfit.
8). Why when analysing with Qdec I get more clusters? Are this defining or representing the sig.mgh as in GLM?
They should give identical results when run in the same way. My
guess is
that you have created two different designs and so are getting
different
answers.
9). When using Qdec were I can find as an output of results the number of clusters and the cortical thickness value?
You'll have to run the correction for multiple comparisons
(interface on
the results page). This simply runs mri_glmfit-sim.
doug
Thank you and have a great day! Antonella
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu
mailto:Freesurfer@nmr.mgh.harvard.edu
<mailto:Freesurfer@nmr.mgh.harvard.edu
mailto:Freesurfer@nmr.mgh.harvard.edu>
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu mailto:greve@nmr.mgh.harvard.edu
<mailto:greve@nmr.mgh.harvard.edu mailto:greve@nmr.mgh.harvard.edu>
Phone Number: 617-724-2358 Fax: 617-726-7422
Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu
mailto:Freesurfer@nmr.mgh.harvard.edu <mailto:Freesurfer@nmr.mgh.harvard.edu mailto:Freesurfer@nmr.mgh.harvard.edu>
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
The information in this e-mail is intended only for the person to
whom
it is addressed. If you believe this e-mail was sent to you in error
and the
e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent
to you
in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.
-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu mailto:greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422
Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu mailto:Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu mailto:Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu mailto:greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422
Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html
Dear Doug,
Can you please give me some details about how the corrections by multiple comparisons are done? Which step in the following code is doing the corrections?
Are the corrections for multiple comparisons done by cache or by cwpvalthresh or the corrections are done by some parts that are implemented in the simulator (the Monte Carlo simulation)?
I am trying to better understand this simulator coding.
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --cwpvalthresh .025 \ --overwrite
Thank you very much. AK
Hi Antonella, have you looked at the help (--help) for the mri_glmfit and mri_glmfit-sim functions? If you still have questions after that let me know.
doug
ps. If you're still learning about multiple comparison correction, I would recommend reading the following papers. They discuss validating gaussian random fields using monte carlo simulations
ftp://surfer.nmr.mgh.harvard.edu/transfer/outgoing/flat/greve/hayasaka.2003.ni.1014.clustersize.pdf ftp://surfer.nmr.mgh.harvard.edu/transfer/outgoing/flat/greve/hagler.2006.ni.surfcluster.pdf
Antonella Kis wrote:
Dear Doug,
Can you please give me some details about how the corrections by multiple comparisons are done? Which step in the following code is doing the corrections?
Are the corrections for multiple comparisons done by cache or by cwpvalthresh or the corrections are done by some parts that are implemented in the simulator (the Monte Carlo simulation)? I am trying to better understand this simulator coding.
mri_glmfit-sim \ --glmdir rh.age.glmdir \ --cache 2 neg \ --cwpvalthresh .025 \ --overwrite
Thank you very much. AK
freesurfer@nmr.mgh.harvard.edu
-
Antonella Kis -
Bruce Fischl -
Douglas N Greve