Hey Gang,
Thanks for the suggestion to switch to MVM. While the model I run in that isn't identical to the model I run using 3dLME (because MVM automatically models the interaction of your within subjects variable with covariates) I get a generally similar pattern of results from MVM, except now the F stats match the GLT's that test for the effect of a condition collapsing across other levels. It looks like what was happening in 3dLME was the F test was really the effect of that factor at the first level of the other factor. So for group, it was giving me the group effect within the negative valence level. Any thoughts on why this is happening? Either way, MVM is doing the trick now, but I like the flexibility I have to specify my model within LME.
Thanks,
Rick