Dear Christopher,
I would like to compare pre and
post treatment brain activity in 10
subjects, who were exposed (during scan) to 3 different categories
of
stimuli, presented through visual and auditory modalities
(individually). Stimuli for each subject were presented in a 12-run
block design (6 visual, with 2 runs of each of 3 stimuli categories;
6
auditory, with 2 runs of each of 3 stimuli categories). I wish to
analyze separate, and conjoined (V + A), activation over three time
points (1 pre and 2 post treatment scans).
I think I should first combine the 2 runs of each stimuli category
(for visual and auditory presentations), and then the 2 stimuli
modalities (V + A) for conjoined analysis but I'm having difficulty
doing this with SPM5. I then hope to use repeated measures ANOVA.
Is
SPM5 able to handle such analyses? I would be EXTREMELY grateful
for
any advice you can offer.
Yes, SPM can handle repeated measures ANOVA of this sort I think the
simplest thing to do would be to use the summary statistic approach
and analyze your effects with a simple one-way ANOVA at the between-
subject level. This would entail creating contrasts for
Vpre Apre Vpost and Apost
for each subject. The main effect of Pre vs. Post would be tested
with a contrast [1 1 -1 -1]; - this is the conjoined effect and is
composed
of the simple A [0 1 0 -1] and V [1 0 -1 0] main effects. You could also
test
for a main effect of modality [1 -1 1 -1] and the interaction [1 -1 -1
1].
This approach could be extend to cover the category effect using
Vpre1 Vpre2 Vpre 3 Apre1 Apre2 Apre3 ....
You will have to specify the non-sphericity in this ANOVA because there
is more than one contrast (4 or 12) per subject. Unless you have good
reason to think that the error variance is different for the each effect,
I
would specify equal variances but allow for non-independence (i.e.
correlations) with subject.
To do this ANOVA you need to construct summary statistics or contrasts
summarizing the Vpre, A pre etc. I would do this by concatenating
runs
from each subject and using the appropriate contrast (summing over the
2 runs per category and possibly the 2 post scans) at the within-subject
level It is possible to do a run-specific first level and then
average at a
second (within-subject) level, depending on which you find the
simpler?
I hope this helps - Karl