Dear Eduardo,
I'm not quite following all the details here (particularly " 5 background or conditions, a control and …") but I think the important thing is that at the first level you should form contrasts that represent the quantities of interest (e.g. main effects of factors) and then you should feed these contrasts up to the second-level. In that way the first-level has already done the within-subject comparisons, and it doesn't matter what the timings are within each subject, as they can be totally different.
See the wiki page on the GLM for some details about setting up contrasts, but adapt the ANOVA cases to work at first level (you should be able to fairly easily have EVs that model conditions such as A1B1 and so on, and if you take an example where these individual conditions are modelled, then you can see how to form the appropriate contrasts).
All the best,
Mark
On 23 May 2014, at 03:37, Eduardo Garza <[log in to unmask]> wrote:
> Dear all
>
> I would really appreciate some input from the community about this. I'm sure is an easy thing but somehow I can't manage to wrap my head around it.
>
> We scanned healthy participants in one session/run who experienced blocks of 5.5 min auditory background sound. There were 5 background or conditions, a control and 4 different types (2 x 2 factorial or A1B1, A1B2, A2B1, A2B2). During each auditory condition, at some fixed time the subjects received 20 seconds of pain then 20 seconds rest, 5 times per condition. It is 1 run of 28.83 minutes per subject.
>
> What I want to do is to look at the differences in the PAIN fmri maps within auditory conditions. For this I'm using a Factorial Design. However, although the timings for each condition and pain was the same in all subjects, the condition order was different for each participant.
>
> Is there a way to include condition order for each participant (with timings) before doing second-level analysis?
> In this way I can model PAIN under A1B1, PAIN under A1B2, etc.
>
> I see now that it would have been easier for the analysis to do one run per condition instead of doing everything in one big session. Another choice would be perhaps to use fslroi to split the files? However, given that the TR=2.7 and each condition is 5.5 min, this could complicate things.
>
>
> Thank you in advance
>
> Kind regards
>
> Eduardo
|