> I have two groups, one healthy control (n=18) and another patients
> with AD (n=25). Patients in AD group have scores of a psychological
> test (25 scores). I need to look at correlation between decreases in
> intensity (volume) in AD compared to healthy group and the 25 scores in
> SPM5.
>
> My questions:
> * If it were paired t-test, I would probably us ImCalc to get
> the images showing differences and would correlate them with the
> scores as single subject (the difference images) and covariate of
> interest (the scores). Am I correct?.
Do you have one scan per subject? If so, then I can't figure out how a paired
t test would fit in.
> * Here I have a two sample t-test which shows me some areas of
> decrease in AD compared to control, but how I can correlate these
> areas with the psychological scores of the AD subjects?
If you want to look for correlations between the scores and the preprocessed
data, then you would include the scores as a covariate. You may no longer
want to model the different groups though, because doing this would localise
only those correlations that can not be explained by the grouping (i.e. the
scores are likely to correlate with the grouping - so your stats are likely
to be less sensitive).
My knowledge of the stats part of SPM is pretty poor, so you may want to see
what the statisticians have to say.
Best regards,
-John
|