Hi everyone...

I have a study design where a patient group is brought in for 3 separate scans after treatment, once immediately after, once at 6 weeks, and once at 6 months. The control group is brought in immediately after (placebo) and then at 6 months. The behavioral task in the scanner consists of 5 different stimulus types.

The problem is that the patients we use aren't the most reliable of folks, and we are left with a design where we have a different number of subjects enrolled in each GROUP, and those different number of subjects return for their 6 month (and 6 week, for the patients) scans with differing frequencies. So the design is a 3 x 2 (neglecting stimulus, for now), where I have:

Patient-Immediate = 25
Patient-6WEEK = 13
Patient-6MONTH = 6
Control-Immediate = 12
Control-6WEEK = 0
Control-6MONTH = 9

So, it's pretty ugly, yes? My question is, can I use GroupAna? If I can, I think it's an unbalanced 4-way ANOVA where SUBJECT is random and is nested within GROUP which is fixed, and TIME is fixed and STIMULUS (mentioned above) is fixed... this is a type 3, right?

Or do I use my extremely limited R skills to try to get 3dLME up and running? And will 3dLME be able to handle this?

Obviously, any design will need to include the fact that one patient's three different scans are highly correlated, i.e. repeated measures.

I'd like to test a TIMExGROUP interaction for each stimulus as well as for the contrasts between a couple of stimuli, as well as a TIME effect in the patient group for the same stimuli combinations.

Thank you for any insight...

-Daniel