Re: [Freesurfer] Second order (square) link function for GLM

External Email - Use Caution

Hi Douglas,

Thanks. But I want to see the square root of the age because I am assuming that the brain feature (cortical thickness) is varying in a square fashion.

Do I have to put square root values? If yes then, how can I put negative square roots. Since we have to put demeaned values.

Best, Abhinav

On Fri, 1 Mar 2019, 01:44 Greve, Douglas N.,Ph.D., DGREVE@mgh.harvard.edu wrote:

If you wanted to do a simple analysis looking at the effect of age in a linear fashion, you'd have a deisgn matrix something like 1 agesubject1 1 agesubject2 ...

ie, first column all ones, second column the ages of each subject. If you think the change is quadratic, then you'd use

1 agesubject1^2 1 agesubject2^2 ...

ie, first column all ones, second column the square of the ages of each subject.

Then use a contrast [0 1] to test whether the 2nd column regressor is diff than 0. (btw, you should demean your ages before squaring)

On 2/28/19 12:12 PM, Abhinav Yadav wrote:

...

    External Email - Use Caution

Hi Douglas,

Thanks for your reply. I am not sure how to do this with design

matrix. I am trying to get higher order fit. Basically I want to see that whether the cortical thickness varies in a quadratic fashion with the behaviour score.

Are you are suggesting that I should put square root values of the

behavioural scores in the design matrix? I am not sure if that is the right approach.

Looking forward to hearing back from you.

Abhinav

On Wed, 20 Feb 2019, 04:12 Greve, Douglas N.,Ph.D., <DGREVE@mgh.harvard.edu mailto:DGREVE@mgh.harvard.edu> wrote:

It sounds like you will just need to create your own design matrix
and
then feed it into mri_glmfit with --X (and not include --fsgd)


On 2/12/19 12:03 AM, Abhinav Yadav wrote:
>
>         External Email - Use Caution
>
> Hello,
>
> I am trying to do GLM analysis for cortical thickness with a
> behavioral score. I wanted to explore square fit in the GLM.
Basically
> wanted to see weather the behaviour is related in a quadratic
function
> instead of linear one.
>
> Couldn't find any option for this is GLM fit.
>
>  I was thinking of feeding a squareroot values of the behavioural
> parameter in the GLM as an alternative. But not sure weather
that will
> work. Please any one can help me with this.
>
> Thanks,
> Abhinav
>
> _______________________________________________
> 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

---

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