Dear Eric,
in the following we assume that there is in fact a global signal which
affects all/most voxels to a varying degree. If our experimental design
doesn't induce global changes (e.g. through changes in ventilation) the
expectation of the covariance between the global signal and the
experimental changes would be zero (the actual signal, not the estimate).
However for an actual experiment of finite length the covariance will
typically not be exactly zero, resulting in a biased parameter estimate for
the experimental factor if the global signal is left out of the model.
Best regards Jesper
Jesper Andersson
Wellcome Dept. of Cognitive Neurology
12 Queen Square
London WC1N 3BG
phone: 44 171 833 7484
fax: 44 171 813 1420
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|