Dear Cory,
Concerning the treatment of reaction time, have a look at previous posts
on the topic as well as this:
http://imaging.mrc-cbu.cam.ac.uk/imaging/DesignEfficiency#VII._Should_I_treat_my_trials_as_events_or_epochs_.3F
but, more generally, if you want to remove a regressor for the design
matrix then you can explicitly replace it with zeros after model
specification:
load SPM.mat
SPM.xX.X(:,c) = 0; % where c is the index of the regressor to remove
save SPM.mat SPM -v6
and estimate it in the usual way. There will be some non-estimable
contrasts but it's not an issue here.
Best regards,
Guillaume.
On 09/04/2019 23:49, Cory Coleman wrote:
> In SPM12, we specify a model using ReactionTimes as a parametric
> modulator which is introduced through the pmod structure included in
> Conditions Onsets .mat file
>
>
>
> "In SPM (and possibly others), the program splits the effect into two
> columns - an unmodulated effect and a parametric column - so that the
> researcher can separately estimate the effect of the condition itself
> and of the parametric modulation of that effect. ...
>
>
>
> As with most parametric modulation, the regressor including reaction
> time effects can be separate from the "trial regressor" that models the
> reaction-time-invariant effect of the trial. This corresponds to having
> one column in the design matrix for the condition itself (which doesn't
> have any reaction time effects) and a second, parametrically modulated
> one, which includes reaction times. If your goal is merely to get the
> best model possible, these don't need to be separated (only the second
> of the two, which includes RTs, could go in the model)"
> (per http://mindhive.mit.edu/book/export/html/55
> <https://eur01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fmindhive.mit.edu%2Fbook%2Fexport%2Fhtml%2F55&data=02%7C01%7Cg.flandin%40UCL.AC.UK%7Cc7823785722243d85cb908d6bd3dbb88%7C1faf88fea9984c5b93c9210a11d9a5c2%7C0%7C0%7C636904470178087300&sdata=sgBhaotJMf7ABwHpIXT9LDJ3DFxZ%2BKNIoJsy530n76w%3D&reserved=0> )
>
>
>
> How do I do that? Currently, my design matrix includes both columns:
> the unmodulated condition itself AND the RT-modulated column. How can
> I delete the unmodulated column from my matrix, leaving me with only the
> RT-modulated columns?
>
--
Guillaume Flandin, PhD
Wellcome Centre for Human Neuroimaging
UCL Queen Square Institute of Neurology
London WC1N 3BG
|