External Email - Use Caution
---------- Forwarded message ---------- From: *abedi000@citymail.cuny.edu abedi000@citymail.cuny.edu* < abedi000@citymail.cuny.edu> Date: Thursday, April 11, 2019 Subject: Fw: mriglmfit error To: Amrita Bedi amibedi273@gmail.com
Amrita Bedi *Biomedical Engineering - Class of 2018* *The City College of New York - CUNY*
------------------------------ *From:* abedi000@citymail.cuny.edu *Sent:* Thursday, April 11, 2019 8:38 AM *To:* Freesurfer@nmr.mgh.harvard.edu *Subject:* mriglmfit error
Good morning Freesurfer developers,
While using the mri-glmfit command on 20 subjects. I got the following error-
-------------------------------- ERROR: matrix is ill-conditioned or badly scaled, condno = 1e+08 -------------------------------- Possible problem with experimental design: Check for duplicate entries and/or lack of range of continuous variables within a class. If you seek help with this problem, make sure to send: 1. Your command line: mri_glmfit.bin --y test.10B.mgh --fsgd TEST.fsgd --C TEST.mtx --surf fsaverage lh --cortex --glmdir test.glmdir 2. The FSGD file (if using one) 3. And the design matrix above Attempting to diagnose further SumSq: Min=0.000000 (col 6), Max=228.790298 (col 3) The scale is much different between columns 6 and 3, you may want to normalize by subtracting the mean and dividing by the standard deviation. Column 6, all values are 0.
1. My command line used was-
mri_glmfit --y test.10B.mgh --fsgd TEST.fsgd --C TEST.mtx --surf fsaverage lh --cortex --glmdir test.glmdir
2. My fsgd file and design matrix file are attached.
Thank you for your help.
Amrita Bedi
Hi Amrita, first of all, thanks for following the instructions on submitting the bug report! It helps us a lot to get all the information the first time. As for your issue, there is a confound between the Female class and the NSI_1 variable. All the Females have a NSI_1 value of 0, meaning that you cannot compute a slope for the Females group. What to do next depends a little on what you think is happening in the data. You could run it with different-offset same-slope (mri_glmfit --fsgd TEST.fsgd DOSS ...). This would force a single slope for both males and females. This is not a problem for the NSI_1 variable because there is no slope for females. If you think this is ok for CAPS_C, then just do it. If not, then things get harder, but I can probably step you through it.
On 4/11/19 8:40 AM, Amrita Bedi wrote:
External Email - Use Caution
---------- Forwarded message ---------- From: *abedi000@citymail.cuny.edu mailto:abedi000@citymail.cuny.edu* <abedi000@citymail.cuny.edu mailto:abedi000@citymail.cuny.edu> Date: Thursday, April 11, 2019 Subject: Fw: mriglmfit error To: Amrita Bedi <amibedi273@gmail.com mailto:amibedi273@gmail.com>
Amrita Bedi /Biomedical Engineering - Class of 2018/ /The City College of New York - CUNY/
*From:* abedi000@citymail.cuny.edu mailto:abedi000@citymail.cuny.edu *Sent:* Thursday, April 11, 2019 8:38 AM *To:* Freesurfer@nmr.mgh.harvard.edu mailto:Freesurfer@nmr.mgh.harvard.edu *Subject:* mriglmfit error
Good morning Freesurfer developers,
While using the mri-glmfit command on 20 subjects. I got the following error-
ERROR: matrix is ill-conditioned or badly scaled, condno = 1e+08
Possible problem with experimental design: Check for duplicate entries and/or lack of range of continuous variables within a class. If you seek help with this problem, make sure to send: 1. Your command line: mri_glmfit.bin --y test.10B.mgh --fsgd TEST.fsgd --C TEST.mtx --surf fsaverage lh --cortex --glmdir test.glmdir 2. The FSGD file (if using one) 3. And the design matrix above Attempting to diagnose further SumSq: Min=0.000000 (col 6), Max=228.790298 (col 3) The scale is much different between columns 6 and 3, you may want to normalize by subtracting the mean and dividing by the standard deviation. Column 6, all values are 0.
- My command line used was-
mri_glmfit --y test.10B.mgh --fsgd TEST.fsgd --C TEST.mtx --surf fsaverage lh --cortex --glmdir test.glmdir
- My fsgd file and design matrix file are attached.
Thank you for your help.
Amrita Bedi
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
External Email - Use Caution
Hi Dr Greve,
I can’t do the Fsgd in the DOSS format since it won’t be ok for my other 6 variables. I would appreciate it if you could guide me through it.
Best Amrita
On Thursday, April 11, 2019, Greve, Douglas N.,Ph.D. DGREVE@mgh.harvard.edu wrote:
Hi Amrita, first of all, thanks for following the instructions on submitting the bug report! It helps us a lot to get all the information the first time. As for your issue, there is a confound between the Female class and the NSI_1 variable. All the Females have a NSI_1 value of 0, meaning that you cannot compute a slope for the Females group. What to do next depends a little on what you think is happening in the data. You could run it with different-offset same-slope (mri_glmfit --fsgd TEST.fsgd DOSS ...). This would force a single slope for both males and females. This is not a problem for the NSI_1 variable because there is no slope for females. If you think this is ok for CAPS_C, then just do it. If not, then things get harder, but I can probably step you through it.
On 4/11/19 8:40 AM, Amrita Bedi wrote:
External Email - Use Caution---------- Forwarded message ---------- From: *abedi000@citymail.cuny.edu mailto:abedi000@citymail.cuny.edu* <abedi000@citymail.cuny.edu mailto:abedi000@citymail.cuny.edu> Date: Thursday, April 11, 2019 Subject: Fw: mriglmfit error To: Amrita Bedi <amibedi273@gmail.com mailto:amibedi273@gmail.com>
Amrita Bedi /Biomedical Engineering - Class of 2018/ /The City College of New York - CUNY/
*From:* abedi000@citymail.cuny.edu mailto:abedi000@citymail.cuny.edu *Sent:* Thursday, April 11, 2019 8:38 AM *To:* Freesurfer@nmr.mgh.harvard.edu mailto:Freesurfer@nmr.mgh.harvard.edu *Subject:* mriglmfit error
Good morning Freesurfer developers,
While using the mri-glmfit command on 20 subjects. I got the following error-
ERROR: matrix is ill-conditioned or badly scaled, condno = 1e+08
Possible problem with experimental design: Check for duplicate entries and/or lack of range of continuous variables within a class. If you seek help with this problem, make sure to send:
- Your command line:
mri_glmfit.bin --y test.10B.mgh --fsgd TEST.fsgd --C TEST.mtx--surf fsaverage lh --cortex --glmdir test.glmdir 2. The FSGD file (if using one) 3. And the design matrix above Attempting to diagnose further SumSq: Min=0.000000 (col 6), Max=228.790298 (col 3) The scale is much different between columns 6 and 3, you may want to normalize by subtracting the mean and dividing by the standard
deviation.
Column 6, all values are 0.
- My command line used was-
mri_glmfit --y test.10B.mgh --fsgd TEST.fsgd --C TEST.mtx --surf fsaverage lh --cortex --glmdir test.glmdir
- My fsgd file and design matrix file are attached.
Thank you for your help.
Amrita Bedi
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
In that case, you will have to create your own design matrix. The easiest way to do that is to start with the DODS Xg.dat matrix that you sent. There will be two columns for the NSI_1 variable (one for males, one for females). You need to reduce it down to one column. In the Xg.dat that you sent, the two columns are the last two columns; in this case, you can just remove the last column. Your contrast matrix will then need one less column.
On 4/11/19 2:08 PM, Amrita Bedi wrote:
External Email - Use Caution
Hi Dr Greve,
I can’t do the Fsgd in the DOSS format since it won’t be ok for my other 6 variables. I would appreciate it if you could guide me through it.
Best Amrita
On Thursday, April 11, 2019, Greve, Douglas N.,Ph.D. <DGREVE@mgh.harvard.edu mailto:DGREVE@mgh.harvard.edu> wrote:
Hi Amrita, first of all, thanks for following the instructions on submitting the bug report! It helps us a lot to get all the information the first time. As for your issue, there is a confound between the Female class and the NSI_1 variable. All the Females have a NSI_1 value of 0, meaning that you cannot compute a slope for the Females group. What to do next depends a little on what you think is happening in the data. You could run it with different-offset same-slope (mri_glmfit --fsgd TEST.fsgd DOSS ...). This would force a single slope for both males and females. This is not a problem for the NSI_1 variable because there is no slope for females. If you think this is ok for CAPS_C, then just do it. If not, then things get harder, but I can probably step you through it. On 4/11/19 8:40 AM, Amrita Bedi wrote: > > External Email - Use Caution > > > > ---------- Forwarded message ---------- > From: *abedi000@citymail.cuny.edu <mailto:abedi000@citymail.cuny.edu> <mailto:abedi000@citymail.cuny.edu <mailto:abedi000@citymail.cuny.edu>>* > <abedi000@citymail.cuny.edu <mailto:abedi000@citymail.cuny.edu> <mailto:abedi000@citymail.cuny.edu <mailto:abedi000@citymail.cuny.edu>>> > Date: Thursday, April 11, 2019 > Subject: Fw: mriglmfit error > To: Amrita Bedi <amibedi273@gmail.com <mailto:amibedi273@gmail.com> <mailto:amibedi273@gmail.com <mailto:amibedi273@gmail.com>>> > > > > > Amrita Bedi > /Biomedical Engineering - Class of 2018/ > /The City College of New York - CUNY/ > > > > ------------------------------------------------------------------------ > *From:* abedi000@citymail.cuny.edu <mailto:abedi000@citymail.cuny.edu> <mailto:abedi000@citymail.cuny.edu <mailto:abedi000@citymail.cuny.edu>> > *Sent:* Thursday, April 11, 2019 8:38 AM > *To:* Freesurfer@nmr.mgh.harvard.edu <mailto:Freesurfer@nmr.mgh.harvard.edu> > <mailto:Freesurfer@nmr.mgh.harvard.edu <mailto:Freesurfer@nmr.mgh.harvard.edu>> > *Subject:* mriglmfit error > > Good morning Freesurfer developers, > > > While using the mri-glmfit command on 20 subjects. I got the following > error- > > -------------------------------- > ERROR: matrix is ill-conditioned or badly scaled, condno = 1e+08 > -------------------------------- > Possible problem with experimental design: > Check for duplicate entries and/or lack of range of > continuous variables within a class. > If you seek help with this problem, make sure to send: > 1. Your command line: > mri_glmfit.bin --y test.10B.mgh --fsgd TEST.fsgd --C TEST.mtx > --surf fsaverage lh --cortex --glmdir test.glmdir > 2. The FSGD file (if using one) > 3. And the design matrix above > Attempting to diagnose further > SumSq: Min=0.000000 (col 6), Max=228.790298 (col 3) > The scale is much different between columns 6 and 3, you may want to > normalize by subtracting the mean and dividing by the standard deviation. > Column 6, all values are 0. > > 1. My command line used was- > > mri_glmfit --y test.10B.mgh --fsgd TEST.fsgd --C TEST.mtx --surf > fsaverage lh --cortex --glmdir test.glmdir > > > 2. My fsgd file and design matrix file are attached. > > Thank you for your help. > > > > > Amrita Bedi > > > > > _______________________________________________ > Freesurfer mailing list > Freesurfer@nmr.mgh.harvard.edu <mailto:Freesurfer@nmr.mgh.harvard.edu> > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer <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 <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
External Email - Use Caution
Hi Dr Greve,
Is the idea behind removing one column for the NS 1 column in the design matrix meant to remove the columns for females? And how do we know which of the two columns for the Ns_1 in the design matrix are for which gender?
In the contrast (0 0 0 0 1-1) which column of the last two for the Ns_1 would be removed?
For the DOSS, say it works for the Ns_1 since female scores for Ns_1 are 0 and we can say that the males and females are decreasing/increasing at the same rate. Would it be right to make the same assumption for Caps_c even though the caps score for females or males are not 0?
Best Amrita ---------- Forwarded message ---------- From: *Greve, Douglas N.,Ph.D.* DGREVE@mgh.harvard.edu Date: Thursday, April 11, 2019 Subject: Fw: mriglmfit error To: "freesurfer@nmr.mgh.harvard.edu" freesurfer@nmr.mgh.harvard.edu
In that case, you will have to create your own design matrix. The easiest way to do that is to start with the DODS Xg.dat matrix that you sent. There will be two columns for the NSI_1 variable (one for males, one for females). You need to reduce it down to one column. In the Xg.dat that you sent, the two columns are the last two columns; in this case, you can just remove the last column. Your contrast matrix will then need one less column.
On 4/11/19 2:08 PM, Amrita Bedi wrote:
External Email - Use CautionHi Dr Greve,
I can’t do the Fsgd in the DOSS format since it won’t be ok for my other 6 variables. I would appreciate it if you could guide me through it.
Best Amrita
On Thursday, April 11, 2019, Greve, Douglas N.,Ph.D. <DGREVE@mgh.harvard.edu mailto:DGREVE@mgh.harvard.edu> wrote:
Hi Amrita, first of all, thanks for following the instructions on submitting the bug report! It helps us a lot to get all the information the first time. As for your issue, there is a confound between the Female class and the NSI_1 variable. All the Females have a NSI_1 value of 0, meaning that you cannot compute a slope for the Females group. What to do next depends a little on what you think is happening in the data. You could run it with different-offset same-slope (mri_glmfit --fsgd TEST.fsgd DOSS ...). This would force a single slope for both males and females. This is not a problem for the NSI_1 variable because there is no slope for females. If you think this is ok for CAPS_C, then just do it. If not, then things get harder, but I can probably step you through it. On 4/11/19 8:40 AM, Amrita Bedi wrote: > > External Email - Use Caution > > > > ---------- Forwarded message ---------- > From: *abedi000@citymail.cuny.edu <mailto:abedi000@citymail.cuny.edu> <mailto:abedi000@citymail.cuny.edu <mailto:abedi000@citymail.cuny.edu>>* > <abedi000@citymail.cuny.edu <mailto:abedi000@citymail.cuny.edu> <mailto:abedi000@citymail.cuny.edu <mailto:abedi000@citymail.cuny.edu>>> > Date: Thursday, April 11, 2019 > Subject: Fw: mriglmfit error > To: Amrita Bedi <amibedi273@gmail.com <mailto:amibedi273@gmail.com> <mailto:amibedi273@gmail.com <mailto:amibedi273@gmail.com>>> > > > > > Amrita Bedi > /Biomedical Engineering - Class of 2018/ > /The City College of New York - CUNY/ > > > > ------------------------------------------------------------
------------
> *From:* abedi000@citymail.cuny.edu <mailto:abedi000@citymail.cuny.edu> <mailto:abedi000@citymail.cuny.edu <mailto:abedi000@citymail.cuny.edu>> > *Sent:* Thursday, April 11, 2019 8:38 AM > *To:* Freesurfer@nmr.mgh.harvard.edu <mailto:Freesurfer@nmr.mgh.harvard.edu> > <mailto:Freesurfer@nmr.mgh.harvard.edu <mailto:Freesurfer@nmr.mgh.harvard.edu>> > *Subject:* mriglmfit error > > Good morning Freesurfer developers, > > > While using the mri-glmfit command on 20 subjects. I got the following > error- > > -------------------------------- > ERROR: matrix is ill-conditioned or badly scaled, condno = 1e+08 > -------------------------------- > Possible problem with experimental design: > Check for duplicate entries and/or lack of range of > continuous variables within a class. > If you seek help with this problem, make sure to send: > 1. Your command line: > mri_glmfit.bin --y test.10B.mgh --fsgd TEST.fsgd --C TEST.mtx > --surf fsaverage lh --cortex --glmdir test.glmdir > 2. The FSGD file (if using one) > 3. And the design matrix above > Attempting to diagnose further > SumSq: Min=0.000000 (col 6), Max=228.790298 (col 3) > The scale is much different between columns 6 and 3, you may want to > normalize by subtracting the mean and dividing by the standard deviation. > Column 6, all values are 0. > > 1. My command line used was- > > mri_glmfit --y test.10B.mgh --fsgd TEST.fsgd --C TEST.mtx --surf > fsaverage lh --cortex --glmdir test.glmdir > > > 2. My fsgd file and design matrix file are attached. > > Thank you for your help. > > > > > Amrita Bedi > > > > > _______________________________________________ > Freesurfer mailing list > Freesurfer@nmr.mgh.harvard.edu <mailto:Freesurfer@nmr.mgh.harvard.edu> > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer <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 <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
_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
External Email - Use Caution
Hi Dr Greve,
Is the idea behind removing one column for the NS 1 column in the design matrix meant to remove the columns for females? And how do we know which of the two columns for the Ns_1 in the design matrix are for which gender?
In the contrast (0 0 0 0 1-1) which column of the last two for the Ns_1 would be removed?
For the DOSS, say it works for the caps score since female scores for Ns_1 are 0 and we can say that the ---------- Forwarded message ---------- From: *Greve, Douglas N.,Ph.D.* DGREVE@mgh.harvard.edu Date: Thursday, April 11, 2019 Subject: Fw: mriglmfit error To: "freesurfer@nmr.mgh.harvard.edu" freesurfer@nmr.mgh.harvard.edu
In that case, you will have to create your own design matrix. The easiest way to do that is to start with the DODS Xg.dat matrix that you sent. There will be two columns for the NSI_1 variable (one for males, one for females). You need to reduce it down to one column. In the Xg.dat that you sent, the two columns are the last two columns; in this case, you can just remove the last column. Your contrast matrix will then need one less column.
On 4/11/19 2:08 PM, Amrita Bedi wrote:
External Email - Use CautionHi Dr Greve,
I can’t do the Fsgd in the DOSS format since it won’t be ok for my other 6 variables. I would appreciate it if you could guide me through it.
Best Amrita
On Thursday, April 11, 2019, Greve, Douglas N.,Ph.D. <DGREVE@mgh.harvard.edu mailto:DGREVE@mgh.harvard.edu> wrote:
Hi Amrita, first of all, thanks for following the instructions on submitting the bug report! It helps us a lot to get all the information the first time. As for your issue, there is a confound between the Female class and the NSI_1 variable. All the Females have a NSI_1 value of 0, meaning that you cannot compute a slope for the Females group. What to do next depends a little on what you think is happening in the data. You could run it with different-offset same-slope (mri_glmfit --fsgd TEST.fsgd DOSS ...). This would force a single slope for both males and females. This is not a problem for the NSI_1 variable because there is no slope for females. If you think this is ok for CAPS_C, then just do it. If not, then things get harder, but I can probably step you through it. On 4/11/19 8:40 AM, Amrita Bedi wrote: > > External Email - Use Caution > > > > ---------- Forwarded message ---------- > From: *abedi000@citymail.cuny.edu <mailto:abedi000@citymail.cuny.edu> <mailto:abedi000@citymail.cuny.edu <mailto:abedi000@citymail.cuny.edu>>* > <abedi000@citymail.cuny.edu <mailto:abedi000@citymail.cuny.edu> <mailto:abedi000@citymail.cuny.edu <mailto:abedi000@citymail.cuny.edu>>> > Date: Thursday, April 11, 2019 > Subject: Fw: mriglmfit error > To: Amrita Bedi <amibedi273@gmail.com <mailto:amibedi273@gmail.com> <mailto:amibedi273@gmail.com <mailto:amibedi273@gmail.com>>> > > > > > Amrita Bedi > /Biomedical Engineering - Class of 2018/ > /The City College of New York - CUNY/ > > > > -----------------------------------------------------------
-------------
> *From:* abedi000@citymail.cuny.edu <mailto:abedi000@citymail.cuny.edu> <mailto:abedi000@citymail.cuny.edu <mailto:abedi000@citymail.cuny.edu>> > *Sent:* Thursday, April 11, 2019 8:38 AM > *To:* Freesurfer@nmr.mgh.harvard.edu <mailto:Freesurfer@nmr.mgh.harvard.edu> > <mailto:Freesurfer@nmr.mgh.harvard.edu <mailto:Freesurfer@nmr.mgh.harvard.edu>> > *Subject:* mriglmfit error > > Good morning Freesurfer developers, > > > While using the mri-glmfit command on 20 subjects. I got the following > error- > > -------------------------------- > ERROR: matrix is ill-conditioned or badly scaled, condno = 1e+08 > -------------------------------- > Possible problem with experimental design: > Check for duplicate entries and/or lack of range of > continuous variables within a class. > If you seek help with this problem, make sure to send: > 1. Your command line: > mri_glmfit.bin --y test.10B.mgh --fsgd TEST.fsgd --C TEST.mtx > --surf fsaverage lh --cortex --glmdir test.glmdir > 2. The FSGD file (if using one) > 3. And the design matrix above > Attempting to diagnose further > SumSq: Min=0.000000 (col 6), Max=228.790298 (col 3) > The scale is much different between columns 6 and 3, you may want to > normalize by subtracting the mean and dividing by the standard deviation. > Column 6, all values are 0. > > 1. My command line used was- > > mri_glmfit --y test.10B.mgh --fsgd TEST.fsgd --C TEST.mtx --surf > fsaverage lh --cortex --glmdir test.glmdir > > > 2. My fsgd file and design matrix file are attached. > > Thank you for your help. > > > > > Amrita Bedi > > > > > _______________________________________________ > Freesurfer mailing list > Freesurfer@nmr.mgh.harvard.edu <mailto:Freesurfer@nmr.mgh.harvard.edu> > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer <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 <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
_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
No, the idea is to merge the males and females into one column. I just so happens that the female column is all 0s and the females have a 0 in the male column, so you can effectively merge them by just removing the last column.
You cannot test for a difference between genders on Ns_1 (which appears to be what your contrast is trying to do) since they are merged into a single regressor
On 4/12/2019 10:48 AM, Amrita Bedi wrote:
External Email - Use Caution
Hi Dr Greve,
Is the idea behind removing one column for the NS 1 column in the design matrix meant to remove the columns for females? And how do we know which of the two columns for the Ns_1 in the design matrix are for which gender?
In the contrast (0 0 0 0 1-1) which column of the last two for the Ns_1 would be removed?
For the DOSS, say it works for the caps score since female scores for Ns_1 are 0 and we can say that the ---------- Forwarded message ---------- From: Greve, Douglas N.,Ph.D. <DGREVE@mgh.harvard.edumailto:DGREVE@mgh.harvard.edu> Date: Thursday, April 11, 2019 Subject: Fw: mriglmfit error To: "freesurfer@nmr.mgh.harvard.edumailto:freesurfer@nmr.mgh.harvard.edu" <freesurfer@nmr.mgh.harvard.edumailto:freesurfer@nmr.mgh.harvard.edu>
In that case, you will have to create your own design matrix. The easiest way to do that is to start with the DODS Xg.dat matrix that you sent. There will be two columns for the NSI_1 variable (one for males, one for females). You need to reduce it down to one column. In the Xg.dat that you sent, the two columns are the last two columns; in this case, you can just remove the last column. Your contrast matrix will then need one less column.
On 4/11/19 2:08 PM, Amrita Bedi wrote:
External Email - Use CautionHi Dr Greve,
I can’t do the Fsgd in the DOSS format since it won’t be ok for my other 6 variables. I would appreciate it if you could guide me through it.
Best Amrita
On Thursday, April 11, 2019, Greve, Douglas N.,Ph.D. <DGREVE@mgh.harvard.edumailto:DGREVE@mgh.harvard.edu <mailto:DGREVE@mgh.harvard.edumailto:DGREVE@mgh.harvard.edu>> wrote:
Hi Amrita, first of all, thanks for following the instructions on submitting the bug report! It helps us a lot to get all the information the first time. As for your issue, there is a confound between the Female class and the NSI_1 variable. All the Females have a NSI_1 value of 0, meaning that you cannot compute a slope for the Females group. What to do next depends a little on what you think is happening in the data. You could run it with different-offset same-slope (mri_glmfit --fsgd TEST.fsgd DOSS ...). This would force a single slope for both males and females. This is not a problem for the NSI_1 variable because there is no slope for females. If you think this is ok for CAPS_C, then just do it. If not, then things get harder, but I can probably step you through it. On 4/11/19 8:40 AM, Amrita Bedi wrote: > > External Email - Use Caution > > > > ---------- Forwarded message ---------- > From: *abedi000@citymail.cuny.edu<mailto:abedi000@citymail.cuny.edu> <mailto:abedi000@citymail.cuny.edu<mailto:abedi000@citymail.cuny.edu>> <mailto:abedi000@citymail.cuny.edu<mailto:abedi000@citymail.cuny.edu> <mailto:abedi000@citymail.cuny.edu<mailto:abedi000@citymail.cuny.edu>>>* > <abedi000@citymail.cuny.edu<mailto:abedi000@citymail.cuny.edu> <mailto:abedi000@citymail.cuny.edu<mailto:abedi000@citymail.cuny.edu>> <mailto:abedi000@citymail.cuny.edu<mailto:abedi000@citymail.cuny.edu> <mailto:abedi000@citymail.cuny.edu<mailto:abedi000@citymail.cuny.edu>>>> > Date: Thursday, April 11, 2019 > Subject: Fw: mriglmfit error > To: Amrita Bedi <amibedi273@gmail.com<mailto:amibedi273@gmail.com> <mailto:amibedi273@gmail.com<mailto:amibedi273@gmail.com>> <mailto:amibedi273@gmail.com<mailto:amibedi273@gmail.com> <mailto:amibedi273@gmail.com<mailto:amibedi273@gmail.com>>>> > > > > > Amrita Bedi > /Biomedical Engineering - Class of 2018/ > /The City College of New York - CUNY/ > > > > ------------------------------------------------------------------------ > *From:* abedi000@citymail.cuny.edu<mailto:abedi000@citymail.cuny.edu> <mailto:abedi000@citymail.cuny.edu<mailto:abedi000@citymail.cuny.edu>> <mailto:abedi000@citymail.cuny.edu<mailto:abedi000@citymail.cuny.edu> <mailto:abedi000@citymail.cuny.edu<mailto:abedi000@citymail.cuny.edu>>> > *Sent:* Thursday, April 11, 2019 8:38 AM > *To:* Freesurfer@nmr.mgh.harvard.edu<mailto:Freesurfer@nmr.mgh.harvard.edu> <mailto:Freesurfer@nmr.mgh.harvard.edu<mailto:Freesurfer@nmr.mgh.harvard.edu>> > <mailto:Freesurfer@nmr.mgh.harvard.edu<mailto:Freesurfer@nmr.mgh.harvard.edu> <mailto:Freesurfer@nmr.mgh.harvard.edu<mailto:Freesurfer@nmr.mgh.harvard.edu>>> > *Subject:* mriglmfit error > > Good morning Freesurfer developers, > > > While using the mri-glmfit command on 20 subjects. I got the following > error- > > -------------------------------- > ERROR: matrix is ill-conditioned or badly scaled, condno = 1e+08 > -------------------------------- > Possible problem with experimental design: > Check for duplicate entries and/or lack of range of > continuous variables within a class. > If you seek help with this problem, make sure to send: > 1. Your command line: > mri_glmfit.bin --y test.10B.mgh --fsgd TEST.fsgd --C TEST.mtx > --surf fsaverage lh --cortex --glmdir test.glmdir > 2. The FSGD file (if using one) > 3. And the design matrix above > Attempting to diagnose further > SumSq: Min=0.000000 (col 6), Max=228.790298 (col 3) > The scale is much different between columns 6 and 3, you may want to > normalize by subtracting the mean and dividing by the standard deviation. > Column 6, all values are 0. > > 1. My command line used was- > > mri_glmfit --y test.10B.mgh --fsgd TEST.fsgd --C TEST.mtx --surf > fsaverage lh --cortex --glmdir test.glmdir > > > 2. My fsgd file and design matrix file are attached. > > Thank you for your help. > > > > > Amrita Bedi > > > > > _______________________________________________ > 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 <https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer> _______________________________________________ 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 <https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer>
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edumailto:Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edumailto:Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edumailto:Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu
-
Amrita Bedi -
Greve, Douglas N.,Ph.D.