Dear SPMers,
could you please give me an advice?
we have a group of smokers, who have been scanned twice: at baseline and
under an experimental condition, a PET-neuroreceptor study. There are 2
types of psychometric data available: a) relevant to 2 conditions, b)
irrelevant to a condition. We’d like to correlate PET data with
psychometric data using SPM2.
1) First, we'd like to look at the correlation between scans and psych
parameters (3 covariates for each condition) within each condition. Should
one put all these data (from both conditions) into one model or consider
each condition separately? If the first approach is more valid, what model
should be used and how can one define contrasts?
2) Second, we'd like to correlate changes in PET data with
changes in psych parameters (a). Is the following design correct?
multi-subjects: condition & covariates: 2 cond, 3 covariates, 0
nuisance variables
covariate interaction – no
covariate centering – no centering
global normalisation – no
grand mean scaling – yes
non-sphericity correction – no
contrasts (0 0 1) & (0 0 -1) will show regions, where changes in
receptor availability between 2 conditions are positively/negatively
correlated with changes in psychometric factors (between 2 conditions).
3) Third, we'd like to correlate changes in PET data with psych parameters
(b). Should i use the same design as for 2) but introduce each
subject score twice?
Is there a possibility to implement implicit masking in this model?
Yours,
Igor
|