Dear SPM-listmates,
I am trying to modell a stimulus which was applied constantly
(permanently & at the same intensity) during sessions of 15 shots (TR 1
or 3 s, respectively) over 20 to 30 sessions. (stupid, hey?) Modelling
by boxcars (or cos with 1 basis function) builds up a design matrix for
which beta is not uniquely estimable with just the regressor being
plotted. However, I can go for the calculation and get results, not too
far off the a-priori hypothesized regions. How much is it worth (facing
my "grey" design matrix)? I thought to try to model some attenuation of
the response (which is evident): would that affect statistical
precision?
TIA-
andreas
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|