Dear Robin,
On 08/01/16 09:20, Robin Shao wrote:
> Dear experts,
>
> I have some confusions about what each of the following eigenvariate
> adjustment methods means:
>
> 1) Don't adjust (I guess this is just doing nothing to the data);
>
> 2) Adjust with respect to an F contrast (My understanding is it removes
> the effects of regressors that are in the original design matrix but are
> not in the F contrast. However it may also be that it removes all
> variances that are not accounted for by the regressors in the F
> contrast. Which, if any, of these is correct?)
The F-contrast to specify is the one testing for all the effects of
interest - responses are adjusted by removing variance that can be
predicted by its null space, see spm_regions.m.
> 3) Adjust for everything (this is the most confusing one to me. Can
> anyone tell me what this does?)
This will regress out the entire design matrix, ie you obtain the
residuals of the model (a projection of your data to the space
orthogonal to the one spanned by all the columns of the design matrix).
Best regard,
Guillaume.
> Many thanks in advance!
>
> Robin
--
Guillaume Flandin, PhD
Wellcome Trust Centre for Neuroimaging
University College London
12 Queen Square
London WC1N 3BG
|