________________________________________
From: SPM (Statistical Parametric Mapping) [[log in to unmask]] on behalf of Nicholas Bowman [[log in to unmask]]
Sent: 22 July 2010 20:44
To: [log in to unmask]
Subject: [SPM] Parametric Modulation with only activations
>Is there a mathematical reason that parametrically modulated HRFs have to have some trials that deactivate >(given that this doesn’t necessarily make sense at the biophysical level)?
Yes, all parametric modulations are mean-corrected, so that they are orthogonal to the main effect (ie main regressor that is modulated). However, this does not imply "deactivations" (relative to baseline).
If the data are true activations whose peak varies with the modulator, the parameter estimate for the main effect will be positive enough that the negative values in the parametric modulator simply reduce the peak response relative to its mean response, but these can still above baseline.
In short, it is the sum of the main regressor and the parametric modulator(s) (weighted by their parameter estimates) that fit the data, so interpretation needs to take into account both.
BW,
Rik
|