Hi there SPMers,
I've got a question about two different approaches to running a second-level contrast, and I'm hoping someone can help:
I've been running a group contrast between two conditions (Task A and Task B) by using a t-contrast at the first level to subtract one condition from the other (Task A > Task B), and then putting all of my subjects' resultant con files into a one-way t-test at the second level. I was recently told that an alternative approach might be more sensitive to activations and might be better overall: Running a contrast of each condition minus implicit baseline for each subject, and putting each subject's TWO resultant con files into a paired t-test at the second level, thus canceling out the implicit baseline and yielding a group contrast of Task A > Task B, just as before. I was told that this paired t-test approach would, by virtue of having a larger number of degrees of freedom, be more sensitive, among other potential advantages.
When I run the paired t-test approach, though, the group-level activation map looks very different (not a lot of overlap) from the one-way t-test's map, something that a difference in dfs alone can't explain. Here's a snapshot of the activation map, with the one-way t-test in red and the paired t-test in yellow: https://files.me.com/drumlizardo/7pilvn. Any ideas about what's going on here? Thanks a lot in advance!
James
|