Hi Donald- thank you for your reply.
The activation maps at the second level show the brain completely active (3 clusters large- each t value =70).
The first level activation maps look very sparse- very few participants had 1st level results, those that did- there wasn't all that much there.
We had 15 subjects- and they did a cyberball task (participants play a ball throwing game with a few others- after several throws they are excluded from the game and no longer receive any ball throws). We modelled it as an inclusion block, and an exclusion block. We are interested in the difference in activity between the exclusion period and the inclusion period- and then want to eventually compare that difference between patients groups. Right now we are just trying to get the "main effect" of exclusion-inclusion for the control group.
We checked the data at each step (esp the preprocessing- everything looks ok) and tried running the analysis on subsets of subjects and on different patient groups and still get the same results (whole brain active).
For second level analyses (RFX), if it pulls out consistency between subjects and calculates the beta values/error (within and between subjects) compared to zero (one-sample)- what if there is consistently nothing (or close to) in all areas at the first level for all people? In other words, what if the con files generated at the first level have small beta values/small variance (say all subjects had nothing) = high t value-how would this effect second level results?
Thanks again so much!!!
Rachael
|