Well, when reconstructing the overall time course this should equal to beta for condition regressor * corresponding predicted time course + beta for time modulation regressor * corresponding predicted time course. For that purpose the scaling should not matter, because an e.g. down-scaled predicted time course results in an correspondingly up-scaled beta. But maybe there are instances in which this does not hold.
I was just curious because I have recently been working with a script from http://wagerlab.colorado.edu/wiki/doku.php/help/fmri_help/fmri_statistical_models/parametric_modulation , which creates a "duration modulator" orthogonalized to the condition regressor, with a parametric modulator and a time modulator being the input. But I think I have to look at some data to see whether results are affected or not.
In any case, which lines should be removed if onsets are specified in units of seconds? Or is it sufficient to just delete the "* job.timing.RT / 60"? As far as I could see there are no additional steps using these parameters in that m file.
Best,
Helmut
|