In the first command you only used 5 samples in your simulation. If you read the "Notes" section below the command, you'll see that this should be something like 5000. Only 5 was used because it would take very long to run. In general, you don't need to run the simulation because it has already been run and the data tabulated (this is what is accessed in the 2nd command line).

On 03/02/2016 10:03 PM, Karl Liu wrote:

Dear Freesurfer experts:

I ran into a confusing problem when applying Monte Carlo simulation to control for multiple comparisons. I am using the buckner data from tutorial dataset.

I first tried the simulation based on Freesurfer;s 2012 manual:

(http://www.freesurfer.net/pub/docs/tarballs_for_courses/FreeSurferTutorial-2...)

mri_glmfit-sim \ --glmdir lh.gender_age.glmdir \ --sim mc-z 5 4 mc-z.negative \ --sim-sign neg --cwpvalthresh 0.05\ --overwrite

This produced a "mc-z.negative.sig.cluster.summary". In this text, no clusters survived correction.

Later on, I applied on a slightly different script, based on Freesurfer tutorial

(http://surfer.nmr.mgh.harvard.edu/fswiki/FsTutorial/GroupAnalysis#Clusterwis...)

mri_glmfit-sim \ --glmdir lh.gender_age.glmdir \ --cache 4 neg \ --cwp 0.05\ --2spaces

It produced a "cache.th40.neg.sig.cluster.summary" this time. But according to this result, 22 clusters survived.

Now I am quite confused, because both corrections have the same voxel-wise (p=0.001) and cluster-wise (p=0.05) threshold. Where the drastic discrepancies come from and which result is more reliable?

Many thinks in advance.

Best,

Karl

---

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