> Dear SPMers,
>
> I have performed a fast-rate event-related fMRI study, in which there are
4
> trial types per run (about 20 trials per condition).
>
> The first trial of each run, however, is not one of the 4 trial types
above.
> I'm wondering what would be the best way to deal with it.
>
> I could:
>
> (1) Not code the first trial (in which case its variance is accounted for
by
> the baseline and/or immediately following trial types in the run)
>
> or
>
> (2) Code the first trial as a completely different trial type (in which
case
> a bad estimate of the first trial's response might contaminate the
estimates
> for other trial types, but perhaps also prevent variance from the first
> trial from affecting the baseline and other trial types to a greater
degree
> than is the case if the first trial was not coded as in option 1 above).
>
>
> I've actually performed random effects analyses across 15 subjects (in the
> context of a finite impulse response model) using both methods of coding
> above.
>
> The results vary in some regions with the method of coding even though
only
> 1 trial (out of 80 per run) is involved!
>
> Would anyone be able to advise me on which method of coding is more
correct?
>
> Thanks
> : ) Daniel
>
> Daniel Weissman
> Center for Cognitive Neuroscience
> Box 90999
> Duke University
> Durham, NC 27708
> phone: 919-681-1029
> fax: 919-681-0815
>
|