[Marsbar-users] percent signal change for parametric modulations

[Jan G. <gla...@hs...>]

Dear Matthew and other MarsBar experts,

I have posted the following email to the SPM mailing list last week, but
did not get a response. So since this is (indirectly) related to MarsBaR I
thought you could give me some comments ...

Thanks a lot for giving it some thought,
Jan

--- Forwarded Mail --------------------------------------------------------

Dear SPM experts,

I know, I know, percent signal change (psc) and how it is computed has been
discussed over and over on this list. However, here is another twist to it
and I wanted to get your opinion on it ...

Generally, one should compute a psc according to this recipe:
http://marsbar.sourceforge.net/faq.html#percent_signal

that is:
psc = (beta(task) * max(regr) * 100) / beta(constant)

where beta(task) is the effect of interest, max(regr) is the maximum of a
new convolved regressor, and beta(constant) is the session constant. The
newly convolved regressor contains only a single event with the duration of
the actual experiment and is convolved with the actual basis function
(which could be an HRF, but also one of the more complex basis sets).

However, when I have a parametric modulation as my effect of interest, then
the corresponding onset regressors models the average activation to all
events, and the parametric modulator models only the deviation of each
event from that average activation.

This implies, I *think*, that in order to get the correct psc one would
have to add the average activation on top of the effect of the modulator.
Or in an equation:

psc = ((beta(ons) + beta(pm)) * max(regr) * 100) / beta(constant)

where beta(ons) is the onset regressor that models the average activation,
beta(pm) is the beta of the parametric modulator and the rest is as above.

Does this all make sense?

Thanks,
Jan


-- 
Jan Gläscher, Ph.D.         Div. Humanities & Social Sciences
+1 (626) 395-3898 (office)  Caltech, Broad Center, M/C 114-96
+1 (626) 395-2000 (fax)     1200 California Blvd
gla...@hs...    Pasadena, CA 91125