Dear Jan,
Here it is. In spm_spm, the inverse is computed (2nd line below) and
is based on the rank of the matrix, hence the reduction.
Satra
%-Design space and projector matrix [pseudoinverse] for WLS
%=======================================================================
xX.xKXs = spm_sp('Set',spm_filter(xX.K,W*xX.X)); % KWX
xX.pKX = spm_sp('x-',xX.xKXs); % projector
and later ...
beta = xX.pKX*KWY; %-Parameter estimates
On Wed, 23 Mar 2005 17:43:16 +0100, Jan Gläscher
<[log in to unmask]> wrote:
> Dear Satra, Bas and list,
>
> I didn't know that SPM2 does that for you and I couldn't find it in the
> code. AFAIK, spm_get_ons creates a stimulus matrix u fomr the specified
> onsets and spm_Volterra convolves this stimulus matrix (a collection of
> delta (stick) functions) with the selected basis set to yield the design
> matrix X. After that only filtering and whitening is applied to it.
> What am I missing here? Could you point me to this design matrix
> reduction in the source code?
>
> Thanks, Jan
|