> I do have some missing data that I would potentially like to incorporate, so I am still interested in this analysis using 3dlme.
In that case, 3dLME or 3dLMEr would be a better choice.
> However, I am curious why you would generally suggest 3dMVM instead of 3dLME?
3dMVM would be more suited for a factorial structure with within-subject factors. For example, with two or more within-subject factors, a repeated-measures ANOVA can be more accurately captured through 3dMVM. The LME specification is more powerful for hierarchical/nested data structures.
> Should I identify contrast active voxels by combining the clusters from Ex and Rest glt, or maybe the F Condition output?
The F-statistic should be similar to what you get from the GLT specification in terms of statistical evidence. However, the former does not provide the directionality information (positive or negative). So, you do need to get the latter for that.
Gang