Thanks Anderson.
I am definitely including a scanner covariate in my group analyses (just 2nd level) to regress out any influences from multiple scanner usage.
Additionally, for my linear signal increase analysis, I ran a 'side-analysis' to test for any difference between scanners in each group alone. Here I included binary EVs for each group for each scanner (4 total: 2 per group for each scanner). I ran F-Tests to see if there was any significant differences, and lots of activation showed up. Is it necessary to run/use this 'side-analysis' or would it be sufficient to just use the singular scanner covariate in my between-group analysis, as it will regress out scanner variability and show regions that aren't 'necessarily' associated with scanner variability?
Indeed, each subject's time series is acquired on a singular scanner, some with multiple time points, but we're only including one scan per subject. My design setup uses ME: Flame 1. Our subject pool equates to 60 participants: 39 Ctrl, 21 Patients. Would it be recommended to try Flame1+2 in this situation, or would my current setup be sufficient enough? With not a large group pop size, I wouldn't want to sacrifice too much power if possible, especially with this given model design.
Best,
Kevin Japardi
|