The statistical modelling part is not really my area of expertise. I know
what the design matrix could look like (the analysis could be done in many
different ways), but I don't know which buttons you would press to produce
it.
A paired t-test like analysis should give the same generalisability to the
population as statistical testing the difference images. i.e. the variance
estimate is purely inter-subject. You therefore don't need to use a two
level analysis in order to generalise.
Your model would be like a paired t-test design, except that you would use the
values of your covariate and zeros - rather than ones and zeros - in order to
model differences over time. This model would assume that the time between
scans is the same for all subjects.
Best regards,
-John
On Friday 11 February 2005 21:06, Maria Densmore wrote:
> I am working on a VBM analysis (SPM2) with one subject group
> scanned at 2 time periods.
>
> I would like to correlate gray matter change from Time1 to Time
> 2 with a particular covariate. I think I would need to run a
> regression analysis and feed in the difference maps of Time1
>
> >Time2[1 -1] in the form of com.img files for each subject
>
> along with a vector of covariates pertaining to each subjects
> difference map. I'm not sure how to get the con.img file for
> each subject in a VBM analysis though.
>
> Thanks in advance for you help and time.
>
> Maria
|