Hi Jesse,
> 1) is 3dLME the right choice for this, or is 3dMVM just as good?
For this specific case of yours, both programs would be able to handle it. However, 3dLME is a little more flexible. Specifically, 3dMVM would model most of the interactions even though you might not be interested in some of them.
> 2) does the model and glt look correct? I've tried running this with or without the "time : 1*time1 -1*time2"
> portion of the GLT and get the exact same results with/without it.
It looks like that both quantitative variables do not vary within subject. So the correct random effect specification is
-ranEff '~1' \
The GLT coding seems fine.
> 3) does the combined 3dICC_REML + LME approach sounds statistically legit?
When you compute ICC, I assume that you would not set 'time' as fixed effect in 3dICC_REML.
Gang
