 |
 |
Dear André, In case you're interested in the average effects I'd go with appropriately weighted con vectors on single-subject level right from the beginning, that is Diff. contrast visit A: 1/4 1/4 -1/4 -1/4 0 0 0 0 0 0 0 0 1/4 1/4 -1/4 -1/4 0 0 0 0 0 0 0 0 Diff. contrast visit B: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1/4 1/4 -1/4 -1/4 0 0 0 0 0 0 0 0 1/4 1/4 -1/4 -1/4 0 0 0 0 0 0 0 0 As long as everything is scaled consistently it won't affect the statistics of the group models, but at some later point you or someone else might want to plot some estimates from the models, and then you wouldn't look at the arithmetic mean but at the sum. Based on these two con images you could then test for changes over time and interactions group x time within a Flexible factorial, see attachment. You have to specify the conditions accordingly, thus e.g. [1 1 1 2] for subjects from the 1st group with first column corresponding to group and second column corresponding to time, thus [2 1 2 2] for subjects from the 2nd group. See attachment (sorry, forgot to add it before). Concerning main effect group, you would go with 1/8 1/8 -1/8 -1/8 ... 1/8 1/8 -1/8 -1/8 ... 1/8 1/8 -1/8 -1/8 ... 1/8 1/8 -1/8 -1/8 ... , thus average across conditions within session (~ time) and across session (~ time) and forward this into a two sample t-test, which allows to test whether (G1 A + G1 B)/2 differs from (G2 A + G2 B)/2 and also to assess average activation across groups via [0.5 0.5]. Given the differential contrasts on single-subject level you could also turn to a more complex model with another factor "condition" (one level corresponding to those with positive signs and the other to those with the negative signs in the first two vectors), which would allow to infer whether any of the detected differences trace back to differences during the "control" conditions or the "task" condition. Best Helmut