Yes, we had been talking about this in https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=spm;71399e6e.1402 .
A duration of 0 s for the standard regressor is not really useful in my context. The variable-duration modulator would also catch up effects due to longer duration as such, so rather testing epoch-based models (say, an average of 300 ms) vs. stick functions. But I'm interested in the effect of different reaction times relative to the average RT, not really whether any of the models is superior.
Hm, different conditions will have different average RTs, this is still imperfect for comparison then. With a standard parametric modulator affecting the amplitude one can combine the betas of the modulator and the average response, thus simulating activations for a certain RT and trying to use this as a correction. Not sure about the variable-duration modulator.
In any case it should be possible to compare different variable-duration modulators as they should always reflect changes due to changes in duration/when increasing the duration in units of 1 s.
|