Dear SPMers,
I am a little lost on how to set up the proper statistics for a
longitudinal VBM study and am hoping someone here can provide some
insight. I am conducting a longitudinal VBM study with a baseline and
two follow-ups, each one corresponding to a different drug
intervention. Let's call them baseline (B), Drug X (X), and Drug Y
(Y). The study is in a crossover design with random ordering, so each
subject would get one of the following:
B --> X --> Y
B --> Y --> X
What I'm interested in is the possible effect on brain structure due to
X and due to Y. I've already used HDW as outlined in several other
posts to compute difference maps from X-->B and Y-->B. Now let's say I
want to know how Drug X is different than baseline. I could do a
one-sample t-test on the X-->B maps, but half of the subjects would have
already taken Drug Y. For them, if the effects of Drug Y persist, some
of the changes from X-->B could in fact be due to an effect from Drug
Y. How do I set up a model that not only accounts for which drug the
subject is taking, but also the order of the subject's scans? I could
add a covariate vector representing scan order, but is a simple
regression enough to sufficiently account for this effect? From my
admittedly very basic understanding of statistics, I would guess a
repeated measures ANOVA would be the way to go, but I don't know if
that's right (or even how to set it up in SPM).
Any help would be much appreciated.
Regards,
Neil
|