Hello,
Demeaning your variables and not your data will lead to problems. The global mean needs to be accounted for, either in the model with a column of ones or by demeaning your data with the -D option. An exception to this is if the data already has a zero mean ( on a per-voxel basis ).
Hope this helps,
Kind Regards
Matthew
> On 14 Nov 2016, at 17:11, Mark Wagshul <[log in to unmask]> wrote:
>
> I am hoping that someone can clarify how the demean option for randomise works. I am doing a voxelwise correlation for DTI data with 5 covariates, and have already demeaned all of my EV's, but not the data. I am assuming that I do NOT have to use the -D option to demean the data, and that this would only be needed if the EV's were not demeaned. It seems that when I do add the -D option the analysis goes much slower than without it (I have over 200 datasets, so the only reason I can think of for the slow performance is that it is deameaning on every permutation, which doesn't really make sense to me). Any advice for the best way to run these analyses would be much appreciated, thanks.
>
> Mark
|