External Email - Use Caution

Dear Freesurfer support,

My name is Vicente and I have a question for you regarding mri_volcluster. I am interested in using mri_volcluster to correct for multiple comparison after generating a probability distribution function (pdf) via simulation testing. I know mri_glmfit-sim does this however in my study I am analyzing several tracts and therefore doing multiple comparisons that means that after obtaining the clusterwise pvalue I would need to correct for multiple comparisons. Instead of using Bonferroni (too conservative) there is an alternative method called MaxT (https://secure-web.cisco.com/13Z7rv-ZG0oGX2IKry8GaDvm3EW4QqaSUDpOASnWumxYbf8...). Probably You already know how it works, basically you create a new distribution choosing the most extreme values in each iteration for your different tests (let’s say 20 different tracts). Usually you do this by selecting the largest Tstat in all of the subjects in each iteration ending up with a new probability distribution with the highest Tstats and therefore the p-value you obtain after comparing your original cluster tstat with the distribution can be considered corrected by multiple comparison. However I tried this and some of the p-values obtained where smaller than using the original pdf from mri_glmfit-sim.I would like to understand better how mri_volcluster does the test between the input file and the pdf file so I can create it properly. I gather that the size of the cluster has an influence both in the input and pdf file but I would like to understand better this.PD: Since I am testing different tracts the searchspace of each is different, could this affect also the testing? I know this is a very specific question, please let me know if I didn’t explain myself correctly. I look forward to your answer.

Best, Vicente

Vicente Ferrer Predoctoral Researcher BCBL http://secure-web.cisco.com/1Xh37aa15XLHMt4I6yT8yFkeiIiQU6DNsWruCXUSJJaD1w_9...

Legal disclaimer/Aviso legal/Lege-oharra: http://secure-web.cisco.com/11DVqQesiAlYYXw7EJ5u75OXNizu1qNUooOtomO025GeiRfN...