You might want to consider "regressing out" the effect of scanner at the second level, though with only five in the one group, that might be hard.
If there's only one "group" (i.e., you're not comparing different diagnoses or effects of drug in different subjects), then perhaps you could just ignore the effect of scanner, since you're not doing a subtraction, though to be honest you should put it in any paper you publish.
Best thing as always would be to collect more data, all on one scanning with exactly the same scanning parameters.
|