3dMVM
cannot handle missing data. I recommend using 3dLMEr
or 3dGLMM
instead.
On a separate note, Years_OpioidUse
and Years_StimulantUse
are likely mediators of the age effect. If so, including them when examining age effects may be problematic. For more details, you may find this discussion helpful: Covariate Selection.
Gang Chen