Your design is not necessarily wrong, but your reasoning is not correct. Since these are categorical variables, there are actually two ways that you can set up the model: interaction and non-interaction. The interaction model allows you to test the interaction between your variables (eg, is the difference between A and B affected by race? or does E2 slope differ between A and B?). If there is such a significant effect (regardless of whether you are interested in it or not), you'd need to use the interaction model which would have 6 classes (3 soft factor times 2 race factors). You would then have 2*6 regressors for 12 total. If you wanted to test the effect of E2, you would then create a contrast with 6 zeros followed by 6 ones (or 6 0.166666). Your design (ie, coding each group as a 1 in a given column) is non-interaction model. If that works for you, then it is fine. Sometimes people will run the interaction model, test for effects, if there are no effects, then use the non-interaction model. Typically, this works out well because it is so hard to get effects in neuroimaging:). Your contrast is right for the non-interaction model.
On 6/14/2025 11:08 PM, Muskan Khetan wrote:
External Email - Use Caution
Hi,
I have a question related to setting up the FSGD file for my analysis. Specifically, I’m looking into the effects of hormone levels on the brain in a single group (i.e., not comparing different classes). My model includes both continuous and categorical covariates.
From what I’ve gathered, most examples on the FreeSurfer website focus on group comparisons, which isn't the focus of my analysis. Therefore, I’ve set up an FSGD file with only one class (see the attached screenshot for reference). In this file:
Variables like |soft_A|, |soft_B|, |soft_C|, |child_raceMixed|, and |child_raceWhite| are *categorical covariates*.
Since these are covariates and not grouping factors, I haven’t created separate classes for them.
I believe this approach is appropriate, but I wanted to confirm whether that understanding is correct.
Additionally, regarding the contrast file:
My contrast vector is |0 1 0 0 0 0 0 0 0 0 0|, where the number of elements matches the number of variables in my model (including the intercept).
As I’m specifically interested in the effect of |E2_mean|, I’ve set its position to |1| and all others (including covariates and intercept) to |0|.
Is this the correct way to set up the FSGD and contrast files for this type of single-group analysis?
Thank you in advance for your guidance.
Best, Muskan
Freesurfer mailing list --freesurfer@nmr.mgh.harvard.edu To unsubscribe send an email tofreesurfer-leave@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman3/lists/freesurfer@nmr.mgh.harvard.e...