Dear FSL experts and users,
I’ve been struggling to design an 4x4 mixed model ANOVA for my fMRI data. It looks like FSL is not good with dealing with repeated measures and there are some work-around procedures. However, getting this to incorporate 4 levels is not easy. This is were I need help.
A short summary of the experiment:
We've tested participants from within an age range of 20-80 years and divided them into 4 groups. Each participant had to perform a complex motor task at 4 difficulty levels. What I want to know is if there is any activation group differences in the (assumed and observed) increase in brain activation with an increase in task difficulty. So,if there is a significant interaction between Age group and Task difficulty.
In more detail: We have task 9 runs for each subject in single scan session. Run number 5 is a control run, so there are 8 task runs. In each run subjects perform the the task with 4 levels of difficulty (in increasing order: R1<R2<angle<zigzag), presented in blocks of 6 trials. We have 100 subjects divided into 4 age groups.
This is what I've done so far:
1) 1st level fixed effects analysis in which I contrast the increasing levels of difficulty R1<R2, R2<angle, angle<zigzag. This should give me the activation increase per level of difficulty, per run.
2) A second level fixed effects analysis where I combine the activation per level of difficulty contrasts (from step 1) from all task runs.
3) A third level analysis in which I do an F-test for the effect of age group on the three subject contrasts (R1<R2, R2<angle, angle<zigzag) separately. Something like this (taken from the FSL wiki):
The problem I see now is that I’m not accounting for the repeated measures anywhere. I’m just modeling the fixed effects in the first level contrasts. Second, this is not an all-round F-test, but rather separate F-tests for particular contrasts.
I came across an alternative that may work very well for me, which is the response by Mark Jenkinson to a similar question in 2014.
https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind1406&L=FSL&P=R139441&1=FSL&9=A&J=on&X=471846D97102A7D526&Y=p.vanruitenbeek%40maastrichtuniversity.nl&d=No+Match%3BMatch%3BMatches&z=4
Does this seem like a good possibility?
I very much appreciate any input.
Best regards,
Peter
|