On 07/19/2016 05:11 AM, Caroline Beelen wrote:
Dear FS team,
I would like to correct for multiple comparisons after I run command-line GLM analysis on aparc.stats tables (my ROI’s are few regions from DK-atlas). I tried to do the Monte Carlo simulation but get an error that a masking-file is missing. In general, how could I correct for multiple comparisons when using aparcstats2tables in GLM?
You'll need to use bonferroni or FDR. MC correction only makes sense for a spatial analysis where there is spatial correlation in the data.
Alternatively I performed the analysis in Qdec (excluding a nuisance factor to be able to do so). I tried to implement few ROI’s from DK-atlas, one at a time, such as left fusiform area. However, when running the analysis it seems to take into account the whole area of the brain and investigates its correlation with a particular selected ROI (i.e. left fusiform area). Would this kind of analysis give me the same results as when only performing stats on this particular region?
No, what you are describing is just treating the ROI as a covariate (like age) in a voxel-wise analysis.
__
Actually, for my analysis I’m only interested in few regions of DK-atlas and, if possible, would like to perform it on only these regions, leaving out all of the rest (I’m correcting for ICV anyways). In that case I would also have to correct for fewer multiple comparisons... How could I achieve this in Qdec and/or command-line GLM?
Do a bonferroni correction using the number of regions you are seaching over
__
__
Thanks in advance for your help.
Regards, Caroline
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer