 |
 |
Dear Ruth, I have some doubts how a "normal" regressor would actually affect/disturb the model. If this results in something like 0 0 0 0 1 0 0 0 1 0 0 ... with 1 each time a trial is presented and the subject giving a specific rating, then basically this is a stick function not convolved with the HRF and not taking into account the hemodynamic shift. This should probably affect your regressors of interest, for example the regressor of no interest might actually "steal" some aspects from your regressors of interest. Therefore parametric modulation seems to be more plausible to me (as it is already fixed to the actual trial onset and not to a whole volume), this way you could also test whether the rating has an influence or not. You would still have to decide whether to use a single parametric modulator (taking into account the four different ratings, which might have a linear or non-linear relationship) or four parametric modulators (one for each of the rating categories). Leaving this aside, if you don't expect any influence on the BOLD response due to different ratings why do you want to include such a regressor of no interest? Or is it rather something like pain rating, with higher rating values for example resulting in larger movements? Best, Helmut