Hello ,
Can anyone tell us how to use SPM99 to do a group analysis with fixed-
effect model together with conjunction analysis as suggested by Friston,
Holmes, and Worsley (1999)? I have an fMRI study with a group of 13
subjects. Each subject performed the run as follows:
R A B R A B R A B
here, R is the control task, and A, B are two different tasks.
We did fixed-effect analysis for every subject at firsr-level analysis.
Then, we want to do the second-level analysis to see average effect over
all subjects.
We did a random-effect analysis at first, but the result was not
satisfactory. It almost no activation can be observed(neither p<0.05
corrected for multiply comparsions, nor p<0.001 uncorrected for multiply
comparsions ).
So, we think maybe it'd to do a conjunction analysis with fixed-effect
model.
So we did the analysis steps as follows:
1. fmri models: specify and estimate a model.
2. mum of session: 13
3. select scan for session 1-13: we selected the snr*.img of 13
subjects.
4. interscan interval (secs): 5 (TR=5000ms)
5. are conditions are replicated? Yes
6. are timing/parameters the same? Yes
7. number of conditions or trials: 3
8. name for condition/ trials 1-3: trial1.. trial2.. trial3..
9. SOA: fixed
10. SOA(scans) for trial1: 18
11. times to 1st trial(scans):0
12. SOA(scans) for trial2: 18
13. times to 1st trial(scans):6
14. SOA(scans) for trial3: 18
15. times to 1st trial(scans):12
16. parametric modulation: none
17. are these trials: epochs: fixed response (box-car)
18. convolve with hrf: yes
19. add temporal derivatives: no
20. epoch length for trial1-3: 6.. 6.. 6
21. interactions among trials (Volterra) no
22. user specified regressors: 0
23. remove global effect: scale
24. high-pass filter?: specify
25. low-pass filter: hrf
26. model intrinsic correlations?: none
27. setup trial-specific F-contrast?: yes
At RESULT, we set the contrast as these:
A vs R (-1 1 0 -1 1 0 -1 1 0 -1 1 0 -1 1 0 -1 1 0-1 1 0 -1 1 0 -1 1 0 -1 1
0 -1 1 0 -1 1 0 -1 1 0)
B vs R (-1 0 1 -1 0 1 -1 0 1 -1 0 1 -1 0 1 -1 0 1 -1 0 1 -1 0 1 -1 0 1 -1
0 1 -1 0 1 -1 0 1 -1 0 1)
Our QUESTION is, Is this analysis the conjunction analysis with fixed-
effect model?
And the second question is, Is "corrected for multiply comparsions"
necessary at the conjunction analysis with fixed-effect model? And in fact,
we did the p<0.05 corrected for multiply comparsions.
Thank you very much.
|