Dear Kai,
Although the figure was not attached, the parameter values sound correct. As you say, the parametric regressor is coding for the intensity of the stimulus, which is decreasing across trials. Thus, the parametric values should start out high and end up low, passing through 0 at the middle. When these values are convolved with an HRF, you should get amplitudes that are near 0 for the middle values (i.e., at the middle of the session).
One thing to keep in mind here is that you still have another regressor that codes for the average response to your stimulus. So, even though the values of the parametric regressor are near 0 in the middle, you're still modeling both (1) the average response to your stimulus and (2) how the response to an individual trial deviates from the average response as a function of intensity. For the trials whose parametric values are near 0, you're simply saying that you expect the responses in those trials to be close to the average response.
Hope this helps!
Daniel
Dear list members,
at the moment we are somewhat puzzled by the use of a parameter on the
first level:
we have 30 stimuli of decreasing intensity and model the hrf with an
additional parameter coding the intensity.
Looking at the model (see figure) we conclude, that the parameter seems
to be centered and then convolved with the canonical hrf, thus in the
middle of the session, the modelled amplitudes are near zero. However,
this is not what we would like to test for... am I missing something? Is
there a way to omit the assumed normalization of the parameter, in order
to get decreasing amplitudes thrughout the session?
Thanks for any help!
Kai,
regressors
--
Dr. Kai Lutz
Neuropsychologie
Universität Zürich
Binzmühlestrasse 14
Raum 4.D.03
CH-8050 Zürich
Tel: +41-44-6357395
Fax: +41-44-6357409