Yes, you need to correct over both hemispheres. The way we do this in FreeSurfer is to first correct over the hemispheres independently. This gives a cluserwise p-value for each cluster. We then bonferroni correct across hemispheres, which basically means multiplying the clusterwise p-value by 2. This is what happens when you run mri_glmfit-sim with the --2space flag (or set the clusterwise p-value threshold to .025 when correcting at the .05 level). Did you use either of those techniques?
doug
On 01/27/2014 03:36 PM, Cat Chong wrote:
Dear Experts, I was wondering if you could possibly help me with a reviewer-related question: We conducted a between-group-analysis estimating the interaction between cortical thickness and age. In summary, we wanted to investigate whether the interaction between cortical thickness and age is stronger for one group relative to another (using a DODS design). We cluster-corrected our results using a Monte Carlo cluster correction of p<0.01 and attained significant group differences for several areas over the left and right hemisphere. One of the reviewer commented that although group differences were estimated for each hemisphere separately, they ought to have been cluster-corrected for both hemispheres. Would you be able to tell me whether or not cluster-correcting for both hemispheres is a necessity, or help me understand the possible implications of not having cluster-corrected for both hemispheres? Very Best Regards, Catherine
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer