Dear SPM-munity,
We have a study with:
15 subjects
scanned 3x,
1 covariate/scan
Basically we want to look at the relationship between the covariate and
the scan within each subject and make some sort of population inference
across subjects. What we've tried so far:
1. ignore the between subject factors and treat it as 45 independent
scans with 45 independent covariates and simply run a Basic Models
regression. This seems incorrect since I'd assume this violates i.i.d.
assumptions about errors
2. analyze the data such that we examine all subjects on a given day so
we'd have 15 scans and 15 covariates. This doesn't quite work because
what we really want to get at is a within-subject analysis across time.
I was wondering if we can use the PET Multi-subject: covariates only
model to do this such that we could answer the question that we want
(basically how well does this covariate predict BOLD over time) and do
so in a statistically sound manner.
Any insights would be appreciated.
Regards,
Jejo Koola
|