Hi SPM Users,
I have found a different way to model the group main effects suggested by Glascher & Gitelman (http://www.sbirc.ed.ac.uk/cyril/download/Contrast_Weighting_Glascher_Gitelman_2008.pdf) for my 2x2 longitudinal design. The contrast was:
0.5 0.5 -0.5 -0.5 0 0 0 1/19 1/19 1/19 1/19 1/19 1/19 1/19 1/19 1/19 1/19 1/19 1/19 1/19 1/19 1/19 1/19 1/19 1/19 1/19 -1/22 -1/22 -1/22 -1/22 -1/22 -1/22 -1/22 -1/22 -1/22 -1/22 -1/22 -1/22 -1/22 -1/22 -1/22 -1/22 -1/22 -1/22 -1/22 -1/22 -1/22 -1/22
[The 0.5's are the 4 group*time interaction columns, followed by 3 covariates, and then 19 subjects in group 1 and 22 subjects in group 2]
The contrast was valid, however the results that came back were huge (I used no masking, and a 0.05 FWE), on the order of a 60,000-voxel significant cluster.
When I run time 1 and time 2 in separate t-tests, the effects come back MUCH smaller. Any ideas why there would be such a large discrepancy?
Thanks!
Sara
RE:
I have built a flexible factorial design using CAT12 for two groups with two sessions (G1T1, G1T2, G2T1, G2T2) and I have a few questions about the contrasts to run.
1. In the first model, I had 3 factors - subject, group, and time and did a main effect of subject, and an interaction effect of group x time. I am mostly interested in the interaction, so I ran an F-contrast and I just want to make sure it is actually modeling the interaction! I set it up as the following:
1 -1 0 0
-1 1 0 0
0 0 1 -1
0 0 -1 1
Does that look right?
2. In that same model as in 1, I can get the main effect for time, but SPM is not letting me run a contrast to look at the main effect of group. So the t-contrasts for the time main effect (G1T1 > G2T2 "1 -1 0 0") was valid, but the t-contrast for group main effect (G1T1 > G2T1 "1 0 -1 0" or G1T2 > G2T2 "0 1 0 -1") is always invalid in the contrast manager. Any ideas why that would be?
I also tried a different model, same as above by I added in the main effects for both group and time in the flexible factorial model. I attached the design matrix for that new model as well. The first two rows of the design matrix is just Group, so when I do a t-contrast for group ("1 -1") that is also invalid.
Any help or ideas would be greatly appreciated!
Thanks so much,
Sara Heyn
|