Hi,
I'd run the data through MELODIC, find the activation
component(s) and then look at the time course
associated with that in order to work out the appropriate
timing.
All the best,
Mark
On 18 Mar 2010, at 03:24, Christopher Benjamin wrote:
> Hi FSL gurus
>
> I've been having a discussion with a colleague and we'd appreciate
> some
> listserv input! If you had a legacy dataset where you were
> concerned that
> all event onsets could be out by one TR (for any given session), but
> had no
> way of telling from runtime info (logfiles etc.), how would you deal
> with
> this? Some possibilities seem to be
>
> 1. Analyze data with both onsets, and look at which gives you the
> expected
> pattern
> 2. Use one set of onsets with temporal derivatives, which should
> account for
> any timeshift
> 3. Look at the error associated with the models and for each
> subject/session, take the model with the lower error.
>
> I think 1 could be a little tricky justify if you later wanted to
> publish,
> and am not sure about 2 or 3.. maybe 2 is most defensible? Any
> perspectives?
>
> Thanks!
>
> Chris
>
|