On Wednesday, September 21, 2011 14:59:12 you wrote:
> Dear FSL experts,
>
> I am trying to run Melodic in order to investigate signal in my data that
> is of no interest (activity related to button pushing/sensation).
So is this data of interest or not? If you've identified a component as
artifact and you want to work around it, look at fsl_regfilt.
> I would
> like to run group ICA. However, my regressors for the experiment are based
> on reaction time and therefore different for each subject. How do I go
> about seeing the relationship between components and stimulus time series
> when the time series is different for each subject?
Variability in reaction time is on the order of ~0.2 seconds, whereas your TR
is probably 2 or 3 seconds. So if all your subjects were reacting to the same
task, it's possible your time series isn't as different between subjects as
you think. Otherwise, you may want to do ICA on each subject separately or
else look at backprojections of your putative group IC in each subject
separately (i.e. stage 1 of dual regression).
|