On 6/6/18 7:14 AM, Yann Quidé wrote:

     External Email - Use Caution

Dear all,

Using Freesurfer 5.3, I would like to limit my whole-brain analyses to a mask I already defined (ie, small volume correction-like analysis). To correct for multiple comparisons, I want to use the mri_glmfit-sim option. As I am looking within my own mask, I understand I need to run my own Monte-Carlo simulation. Would 10,000 permutations be sufficient?

Yes, you can probably get away with 1000 even.

Also, I am a bit confused on what is best practice when setting the cluster forming threshold. Should I use the mc-z=3 (that is p=0.001) or mc-z=1.3 (that is p=0.05)?

For mc simulation, you probably need 3 (ie, p<.001). If you need to use a lower threshold because none of your voxels survive, then consider using permutation instead of mcz.

Here is an exemple of what I’m doing so far:

mri_glmfit-sim \ --glmdir $STATS_DIR/lh.my_analysis.glmdir \ --sim mc-z 10000 3 mc-z.abs.3 \ --sim-sign abs

Thanks for your time.

Yann

---

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