> I am concerned that if we have the same stimulus feature
> repeated in a clip - which happens quite often (and in fact more
> often for 2 of the three stimulus features which is unavoidable), we
> will run out of power for decovolution (which as far as I understand
> it assumes linearity). For example, if we have a clip that is 12
> seconds long (6 TR's), and has the same feature in every TR, if we
> mark every TR, we will end up with a 1D file that has six ones in a row.
If power is a mjor concern, how about if you adopt a fixed model such as Gamma instead of deconvolution? Since the stimuli are more or less of block type, a fixed model might not be a bad idea. Or try both, and see how it goes.
Gang