Yes, once you take the difference between two time points, then the steps are the same as cross-sectional analysis. But if you took the difference, you should have only one entry in the FSGD file for each subject (looks like you have 2). Also, it is probably a good idea to normalize the ages (or whatever that first variable is) by removing the mean and dividing by the stddev.

On 2/1/2022 10:31 AM, Edwards, Vincent wrote:

External Email - Use Caution

Good morning Freesurfer Team!

I would like to use GLM on my longitudinal study to measure change between two time points. I followed the longitudinal tutorial, and to my understanding, I would move on to the general glm tutorial next. The longitudinal tutorial says that "you can compare groups similar to a cross sectional analysis", but when we run stats on our significant clusters our results look "odd". I may have misunderstood the tutorial, but I plugged the longitudinal timepoints into my fsgd file like this:

Input *Subject1*_*TP1* subject 57   0.93

Input *Subject1*_*TP2 *  subject    57   0.93

Input *Subject2*_*TP1*   subject    64   1.22

Input *Subject2*_*TP2*   subject    57   1.22

I'm not sure if what we ran is actually finding change between timepoints in our subjects. We happened across a glm paired-analysis tutorial. Should we re-run our data through that instead? The longitudinal tutorial only referenced the regular glm tutorial, so I want to make sure that we are following the correct steps - perhaps there is something I did wrong with my fsgd file? The paired-analysis tutorial seems like it may be appropriate, but I was hoping you could clear up our confusion before we move forward.

Thank you, Vince

---

Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer