On 6/15/14 2:24 PM, wangzhiwei3233 wrote:
Hi, all, I have two quesiotns about clusterwise correction for multiple comparsion. I run 2 corrections for one same result,and commands are as follows: a. mri_glmfit-sim --glmdir g2v0.weight --cache 2.0 *abs* --cwpvalthresh .05 --2spaces b. mri_glmfit-sim --glmdir g2v0.weight --cache 2.0 *pos* --cwpvalthresh .05 --2spaces The only difference is the sign of voxel-wise results.
*Question 1* : Should the postive clusters from "*abs"* be same as the cluster from "*pos*"? If yes, why cluster size from "*pos*" are different from postive cluster from "*abs*" in my case?(see results) *Resluts form a (only positive clusters are listed):* # ClusterNo Max VtxMax Size(mm^2) MNIX MNIY MNIZ CWP CWPLow CWPHi NVtxs Annot 3 6.483 140982 *222.17* -14.8 58.9 -6.1 0.01510 0.01350 0.01670 281 rostralmiddlefrontal 8 5.208 45661 *266.48 * -6.7 42.3 -22.2 0.00450 0.00370 0.00540 454 medialorbitofrontal 9 5.103 84832 *241.36 * -32.0 30.7 -2.3 0.00830 0.00710 0.00950 655 lateralorbitofrontal *Resluts form b:* ** # ClusterNo Max VtxMax Size(mm^2) MNIX MNIY MNIZ CWP CWPLow CWPHi NVtxs Annot 1 6.483 140982 *253.31* -14.8 58.9 -6.1 0.02000 0.01820 0.02180 326 rostralmiddlefrontal 2 5.208 45661 *402.16 * -6.7 42.3 -22.2 0.00080 0.00050 0.00120 727 medialorbitofrontal 3 5.103 84832 *287.61* -32.0 30.7 -2.3 0.01010 0.00880 0.01140 772 lateralorbitofrontal
Usually they would be the same, but they don't have to be if negative and positive vertices are right next to eachother to form a larger cluster. I would not have thought this could happen, but check your data.
*Question 2* : Is the simulation from pre-cached independent from my results waiting for being corrected?
I'm not sure what you mean. The pre-cached data are "independent" in that they were computed without reference to it (as opposed to FDR or permutation). It is the same for t-tests, F-tests, and all parametric statistics.
Thanks a lot!
Wilder 2014/06/15
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer