External Email - Use Caution
Hello Experts,
I've a question about mri_glmfit. I want to investigate the association between thickness and neuropsychiatric syndromes, controlling for age and cognition. Which model below is mri_glmfit performing?
Model 1 neuropsych syndromes ~ age + cognition + cortical thickness or Model 2 cortical thickness ~ age + cognition + neuropsych syndrome
Many thanks, Paul
Model 2 is the only one we have the ability to fit. Model 1 would need some kind of machine learning method
On 2/23/2021 12:37 AM, miracle ozzoude wrote:
External Email - Use Caution
Hello Experts,
I've a question about mri_glmfit. I want to investigate the association between thickness and neuropsychiatric syndromes, controlling for age and cognition. Which model below is mri_glmfit performing?
Model 1 neuropsych syndromes ~ age + cognition + cortical thickness or Model 2 cortical thickness ~ age + cognition + neuropsych syndrome
Many thanks, Paul
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
External Email - Use Caution
Thanks Doug. I appreciate it. I have more questions.
1) Aren't both models answering the same question regardless of which side of the equation cortical thickness or syndromes is placed?
2) Also, I noticed in the mri_glmfit --help, it said that forward model 2 is inverted to solve for the regressor of interest. Correct if I'm wrong, does it mean if my matrix is 0 0 0 1 where 1 represents neuropsych syndrome, it will solve for it?
The forward model is given by:
y = W*X*B + n
where X is the Ns-by-Nb design matrix, y is the Ns-by-Nv input data
set, B is the Nb-by-Nv regression parameters, and n is noise. Ns is
the number of inputs, Nb is the number of regressors, and Nv is the
number of voxels/vertices (all cols/rows/slices). y may be surface
or volume data and may or may not have been spatially smoothed. W
is a diagonal weighing matrix.
During the estimation stage, the forward model is inverted to
solve for B:
B = inv(X'W'*W*X)*X'W'y
3) Lastly, how do I extract the beta-values after running mri_glmfit-sim without matlab?
Many thanks, Paul
On Tue, Feb 23, 2021 at 12:37 AM miracle ozzoude miracooloz@gmail.com wrote:
Hello Experts,
I've a question about mri_glmfit. I want to investigate the association between thickness and neuropsychiatric syndromes, controlling for age and cognition. Which model below is mri_glmfit performing?
Model 1 neuropsych syndromes ~ age + cognition + cortical thickness or Model 2 cortical thickness ~ age + cognition + neuropsych syndrome
Many thanks, Paul
On 2/23/2021 11:37 PM, miracle ozzoude wrote:
External Email - Use Caution
Thanks Doug. I appreciate it. I have more questions.
- Aren't both models answering the same question regardless of which
side of the equation cortical thickness or syndromes is placed?
Similar questions, but the results will not be identical. Eg, if cognition and thickness are identical numbers, then you can't solve Model 1 but you can solve Model 2. I was thinking you wanted to use the thickness from the entire cortex to estimate syndromes.
- Also, I noticed in the mri_glmfit --help, it said that forward
model 2 is inverted to solve for the regressor of interest. Correct if I'm wrong, does it mean if my matrix is 0 0 0 1 where 1 represents neuropsych syndrome, it will solve for it?
Yes, it will. But you will need to create an "image" of your neuropsych scores so that they are in each voxel.
The forward model is given by:
y = W*X*B + n
where X is the Ns-by-Nb design matrix, y is the Ns-by-Nv input data
set, B is the Nb-by-Nv regression parameters, and n is noise. Ns is
the number of inputs, Nb is the number of regressors, and Nv is the
number of voxels/vertices (all cols/rows/slices). y may be surface
or volume data and may or may not have been spatially smoothed. W
is a diagonal weighing matrix.
During the estimation stage, the forward model is inverted to
solve for B:
B = inv(X'W'*W*X)*X'W'y
- Lastly, how do I extract the beta-values after running
mri_glmfit-sim without matlab?
do you want a single beta as an overlay? Then use mri_convert and specify the frame you want with --frame (starts at 0). If you want values, you can run mri_convert beta.mgz --ascii beta.ascii
Many thanks, Paul
On Tue, Feb 23, 2021 at 12:37 AM miracle ozzoude <miracooloz@gmail.com mailto:miracooloz@gmail.com> wrote:
Hello Experts, I've a question about mri_glmfit. I want to investigate the association between thickness and neuropsychiatric syndromes, controlling for age and cognition. Which model below is mri_glmfit performing? Model 1 neuropsych syndromes ~ age + cognition + cortical thickness or Model 2 cortical thickness ~ age + cognition + neuropsych syndrome Many thanks, Paul
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu
-
Douglas N. Greve -
miracle ozzoude