I know this issue has already been addressed in the mailing list but since
no one referred to the codelines in SPM5 I would like to ask the community
again which f-contrast should be selected for adjusting the VOI data. is it
the f-contrast of the effects we are interested in or the f-contrast of
motion regressors etc.
the voi extraction is done by the matlab function spm_regions. if a
f-contrast is selected the following code is executed:
spm_regions.m, line 194
==============
%-Parameter estimates: beta = xX.pKX*xX.K*y
%---------------------------------------------------------------
beta = spm_get_data(SPM.Vbeta,xSPM.XYZ(:,Q));
%-subtract Y0 = XO*beta, Y = Yc + Y0 + e
%---------------------------------------------------------------
y = y - spm_FcUtil('Y0',SPM.xCon(xY.Ic),SPM.xX.xKXs,beta);
==============
in spm_FcUtil there are the following information concerning this method-call
spm_FcUtil, line 458
==============
case 'y0' %- Fitted data corrected for confounds defined by Fc
%=======================================================================
% Y0 = spm_FcUtil('Y0',Fc, sX, b)
[...]
varargout = { sf_Y0(Fc,sX,b) };
==============
spm_FcUtil, line 827
==============
function Y0 = sf_Y0(Fc,sX,b)
Y0 = sX.X*(eye(spm_sp('size',sX,2)) - spm_sp('xpx-',sX)*sf_H(Fc,sX))*b
==============
If I understand the code correctly, the F-Contrast defines the values of Y0.
And since spm_regions computes
Y0 = XO*beta, Y = y-Y0
the f-contrast should be defined upon all regressors and conditions that are
of NO interest. Am I right?
PS: I refer to the code of the standard SPM5 installation and the update 1782.
|