Dear everyone,
a while ago I tried the script from http://wagerlab.colorado.edu/wiki/doku.php/help/fmri_help/fmri_statistical_models/parametric_modulation to create parametric modulators in the context of reaction time modeling that affect amplitude AND duration, in contrast to the default SPM parametric modulator which just affects the amplitude. I encountered some problems at the beginning, but slight adjustments seemed to solve the issue (https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=spm;48ef1a6b.1501 ). The script requires a model in which the condition is specified with onsets and a certain (fixed) duration, a time modulator and a parametric modulator, the latter with the reaction times entered on a trial-by-trial basis.
Now recently I looked at variable-duration modulation again. As I had already specified some conventional models I took some time courses as stored in SPM.xX.X and orthogonalized them with spm_orth(X). Surprisingly, this produces different results.
Thus, assuming I want to construct a modulator also working on the duration: Is it plausible to construct a condition A based on trial onsets and a certain fixed duration (e.g. mean reaction time), to construct another condition B based on the very same onsets and a duration vector reflecting the different reaction times, and then to orthogonalize the 2nd onto the 1st via spm_orth, with the final model being based on the regressor A and the output from spm_orth entered as an extra regressor?
If so, then it would also be interesting to know why results obtained with the approach described above differ from those of the script based on the Grinband 2009 paper (or in other words, what were they aiming for then). If someone's interested in that issue I can provide some data.
Best,
Helmut
|