Hi Liz,
I think the answer to both questions is no, though they aren't
too far off.
There is no de-mean step. Rather the first regressor would be
the response curve at the mean amplitude, and the second would
be the modulation from the mean (and so its mean would be zero).
The regressors are not othogonalized either. However, if the
individual response curves do not overlap (so either slow-event
or block design), I would expect them to be orthogonal. But not
if they do overlap.
Does that seem reasonable?
- rick